Identifying and Interpreting Traditional Architectural Style Characteristics Based on Deep Learning

Identifying and Interpreting Traditional Architectural Style Characteristics Based on Deep Learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Identifying and Interpreting Traditional Architectural Style Characteristics Based on Deep Learning Changyao Chen, Shaodan Li, Jiayou Ding, Yincui Hu, Ziyue Fan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7971684/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 18 You are reading this latest preprint version Abstract From the perspective of Chinese traditional architectural style (Su, Hui, Jing, Min, Chuan and Jin), environmental characteristics, architectural features and architectural landscape characteristics of traditional architectures are identified and interpreted using multi-source data, including remote sensing (RS) images, digital elevation model (DEM) data, online architectural landscape (OAL) images. By deep learning (DL) and clustering analysis methods, this paper constructs an hierarchical framework for feature recognition of Chinese traditional architecture, which systematically reveals the inherent coupling relationship between geographical environment, architectural features, and landscape semantics of traditional architectures. The framework can provide technical support for the protection of Chinese traditional architectural cultural heritage. Architectural style Chinese traditional architectures Deep learning Multi-source data Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Introduction As an important material carrier of regional culture, traditional architectural landscape is a material civilization shaped by natural environment, social-economic foundation, folk culture, and other factors in the long-term historical evolution [ 1 ]. Chinese traditional settlements carry a long history and regional culture. However, with the impact of rapid urbanization, traditional settlements are facing a crisis of survival. On the one hand, standardized construction processes have led to the gradual disappearance of regional characteristics of traditional architectures such as horsehead walls. On the other hand, urban expansion has further damaged the regional characteristic of architectural landscapes. As emphasized by Wu [ 2 ], the cultural value of traditional settlements lies in the complex relationship between the architecture and natural environment, which urgently requires further research. In recent years, there have been numerous studies on traditional settlement landscapes, such as landscape gene recognition, spatial form quantification, and landscape feature extraction. In terms of landscape gene recognition, Liu [ 3 ] took Chinese traditional settlements as research objects, and proposed the “landscape information chain” theory and “cell-chain-shape” analysis model to study traditional settlement landscape gene mining, map construction, and regional division. Hu et al. [ 4 ] constructed a multidimensional traditional settlement landscape genes identification system from spatial form, material elements, and non-material elements, which provided the systematic analysis framework for settlement landscape research. Duan et al. [ 5 ] took traditional settlement buildings in Jiangxi and Anhui provinces as examples to construct the cultural landscape factor system from building types, spatial configuration, and natural environment. They summarized the cultural landscape characteristics and differences in this region. For spatial form quantification, Ren et al. [ 6 ] used GIS technology and landscape pattern index to analyze the layout characteristics of rural settlements using remote sensing images. Wang [ 7 ] integrated spatial syntax theory and structural linguistic methods to explore the systematic nature of traditional settlement space construction through a comparative study of the construction mechanisms of Han and Zhuang ethnic groups in northern Guangxi province. Regarding landscape feature extraction, Li [ 8 ] took the southern region in Shaanxi as a case study and explored a method for identifying and extracting the characteristics and compositional patterns of traditional settlement landscapes in multicultural areas from the settlement to individual buildings. Xing et al. [ 9 ] proposed a landscape heritage information extraction and 3D model based on sketches and text prompts, achieving 3D landscape reconstruction. Lu [ 10 ] proposed a landscape feature extraction model, which evaluates various natural landscape scene data and balances the accuracy and efficiency requirements of remote sensing landscape classification. These studies have promoted quantitative research on traditional settlement landscapes from different regions and levels, providing academic support for understanding landscape characteristics of traditional settlements. The rapid development of deep learning (DL) technology has provided a chance for traditional building recognition. The DL methods for traditional building recognition are divided into image classification, semantic segmentation, and object detection. For image classification, Mathias et al. [ 11 ] used building facade images and street images to identify Flemish Renaissance style, Ottoman style, and neoclassical style buildings, laying the foundation for the automated recognition. Han et al. [ 12 ] used a convolutional neural network (CNN) to classify traditional architectural styles using building facade images. Tan et al. [ 13 ] proposed an automatic classification method for traditional village heritage value elements based on the DL method, taking traditional villages in Hubei Province as the study case. In terms of semantic segmentation, Dai et al. [ 14 ] proposed a facade segmentation model based on residential buildings, which accurately identified the components of building facades using street view images. Haznedar et al. [ 15 ] used PointNet to segment the architectural elements of Türkiye's Gaziantep heritage, and summarized the architectural style differences in different regions of Türkiye. Sun et al. [ 16 ] proposed a weakly supervised semantic segmentation method for ancient architecture based on multi-scale adaptive fusion and spectral clustering, and validated it on the Chinese religious famous mountain ancient architecture dataset and Baroque architecture dataset. For object detection, Xiong et al. [ 17 ] proposed the detection model for traditional Hakka walled houses in China based on ResNet50 and YOLOV2, which achieved excellent performance. Du and Wang [ 18 ] used Bi-YOLO network to identify building components of traditional residences in the southeastern region of Hubei Province. In addition, the technology of feature visualization has improved the interpretability of the research. For example, Obeso et al. [ 19 ] classified the architectural styles using digital photos of Mexican cultural heritage, and demonstrated the representative architectural features that the model focuses on through Class Activation Mapping (CAM). Sun et al. [ 20 ] analyzed the correlation between architectural style and construction years in Amsterdam, and presented the style evolution pattern using t-distribution random neighbor embedding (T-SNE) clustering method. With the popularity of the Internet, many online architectural landscape (OAL) images released by the public have become an important data source for studying traditional architecture. These images are numerous and wide-ranging, featuring both architectural details and architectural environments. Most of them convey representative characteristics of the architectural style. The application of OAL images has also made significant progress. Zu et al. [ 21 ] used DL methods to reveal representative features of remote rural buildings from online building photos of Tibetan and Qiang areas, demonstrating the potential of OAL images in analyzing building and environmental features. Two main limitations exist in extracting architectural features based on OAL images. On the one hand, OAL images contain many irrelevant contents to buildings, such as people, sky, and trees, which interfere with the recognition of architectural features; on the other hand, OAL images focus on the facade features of architectures, lacking roof information of the architectures. Fortunately, the top-down perspective provided by remote sensing images can supplement roof and surrounding environmental information [ 22 ]. Therefore, this study integrated RS, DEM and OAL images to interpret environmental characteristics, architectural features and architectural landscape characteristics of the traditional architectures in terms of architectural styles. The RS and DEM data reveal the natural environment that influences the architectural landscape characteristics. RS and OAL images are used to identify the traditional architectural features and reveal the relationship between these features and the overall architectural landscape. The contributions of this study are as follows: (1) Multi-source data (i.e., RS, DEM and OAL data) provides data support and analytic dimension for the interpretation of traditional architectural style features; (2) This paper innovatively proposes to interpret the traditional architecture characteristics from the perspective of architectural styles; (3) The hierarchical framework “architectural environment characteristics-architectural features།architectural landscape characteristics” of Chinese traditional architecture is constructed,, offering a systematic approach for traditional architectural heritage studies. Data Collection and Processing The traditional buildings used in this paper are selected from the list of villages in the Digital Museum of Traditional Chinese Settlements, and the architectural styles of the buildings in the villages are clearly documented. The location of the traditional villages used in the paper is shown in Fig. 1 . The list of the traditional villages that covers the six architectural styles is given in Appendix A. (1) DEM data: DEM data for each settlement were obtained with a resolution of 12.5m based on the list. These data were used to analyze the topographic and geomorphological features of different architectural styles. (2) RS image: According to the list, RS images were collected using Tuxin Earth GIS tools with a spatial resolution of 0.4 ~ 0.6m. During the collection process, the criteria of complete preservation and no large-scale occlusion were followed. (3) OAL image: In the collection of OAL images, a multi-level image acquisition strategy was adopted: firstly, the county name and settlement name were used as keywords for accurate retrieval on search engines such as Google and Baidu, and it ensures the searched images were concentrated on the target settlement. Then, an extended search using representative architectural styles as keywords was added, effectively improving the representativeness of OAL images. Finally, data cleaning was performed on the collected images. The cleaning standards are given as follows: 1) Retain images that show the complete architectures and surrounding environment; 2) Delete images that lack architectural landscape elements. Table 1 shows the RS and OAL images with different architectural styles and gives detailed descriptions. In this paper, the dataset of traditional architectures was constructed. It consists of two parts: (1) The architectural style classification dataset includes RS and OAL images. The images were matched each other in each village among different architectural styles. Table 2 shows the number of RS and OAL images with each architectural style. (2) Semantic segmentation dataset of architectural landscapes only includes OAL images with each architectural style. Table 2 Statistics of CTS image Style Number of CTS Number of RS images Number of OAL images Su 40 334 334 Hui 40 334 334 Jing 40 334 334 Min 40 334 334 Chuan 40 334 334 Jin 40 334 334 To facilitate the segmentation of architectural landscape elements, the content of OAL images is divided into two classes, that is, natural and artificial landscape elements. A total of 14 landscape element categories are shown in Fig. 2 . Natural landscape elements include the sky, mountains, vegetation, and water. The sky is the background of the image, and its light changes have a significant impact on visual perception. It occupies a large region in OAL images and is more likely to affect the identification of architectural style. Mountains and water are important natural elements that influence the spatial distribution of traditional settlements [ 23 ]. Mountains and water jointly shape the classic settlement pattern of “nestled against mountains and beside rivers” in space. Vegetation is one of the most common landscapes in settlements and architectures. Artificial landscape elements include roads, stone steps, stone balustrades, bridges, gate piers, censers, vats, stone mills, and wells. Gate piers are located on both sides of residential gates, and their morphological characteristics carry ritual information. Censers are also symbolic of architectural etiquette. Production and living utensils, such as vats, stone mills, and wells, meet daily needs and become unique landscapes in each architectural style. Methods Research Framework As shown in Fig. 3 , the overall workflow of this study includes four main steps: data collection, analysis of architectural environment characteristics, interpretation of architectural features and interpretation of architectural landscape features. 1. Data collection: Based on the list of traditional settlements, the RS, DEM, and OAL images of different architectural styles were collected. 2. Analysis of architectural environment characteristics: Using RS and DEM data, the natural environment characteristics of the architecture for six architectural styles were analyzed. 3. Interpretation of architectural features: Based on RS and OAL images, a double-branch network model was employed for architectural style classification. The heatmap generated by the Grad-CAM method is used to visualize the image regions on which the model focuses on. By counting the frequency and quantity of the highlighted regions, the characteristics of the architectural styles can be clearly presented, such as architectural components, roof types, architectural colors, and architectural materials. 4. Interpretation of architectural landscape features: Based on the OAL images, a semantic segmentation model was employed to divide the images into 14 landscape element categories, which include natural and artificial landscape elements. For each architectural style, firstly, the semantic segmentation results are superimposed with the heatmap to obtain the landscape elements that the model is most concerned about. Then, the six elements with the highest frequency of attention are regarded as representative landscape elements of this architectural style. Finally, T-SNE analysis was conducted on the representative landscape elements, and the landscape correlation characteristics of each architectural style were analyzed. Deep Learning Models Double-branch Classification Network Based on the OAL and RS images of traditional architectures, a double-branch network was designed for the classification of the architectural styles. As shown in Figure.4, the proposed model consists of feature extraction and feature fusion. In the feature extraction, a DenseNet backbone was employed for both branches to facilitate robust feature learning [ 24 ]. Specifically, the RS branch specializes in learning the architectural roof morphology and overall layout features, while the OAL branch focuses on the landscapes of the architectures, door and window styles, the texture and materials of the walls. In the feature fusion, the CBAM mechanism was introduced. This mechanism utilizes dual weighting of channel and spatial attention to automatically focus on the most discriminative building features [ 25 ]. In addition, the module also incorporated a feature reuse mechanism to achieve cross branch fusion of high-level features from the OAL branch. For a detailed introduction of the model, please refer to the paper of Zhang et al. [ 26 ]. Semantic Segmentation Network To perform architectural landscape semantic segmentation on OAL images, SegFormer was employed. SegFormer is an advanced semantic segmentation model based on the Transformer architecture [ 27 ], which adopts a combination architecture of layered Transformer encoder (MiT) and lightweight multilayer perceptron (MLP) decoder. The input image is transformed into a multi-scale feature representation through the overlapping block embedding, which can simultaneously capture the detailed features of the building and the overall spatial layout relationships to achieve accurate segmentation [ 28 ]. The network architecture of the model is shown in Figure. 5. Figure 5 shows the framework of the SegFormer. The encoder part adopts a hierarchical design, including multiple stages of Transformer modules. Each stage constructs a multi-scale feature pyramid from high-resolution details to low resolution semantics by gradually reducing sequence length and expanding channel dimensions. For an image with the input size of H × W ×3, SegFormer first uses the overlap patch embeddings technique to divide the image into multiple patches of the size 4×4. Pre-trained CNN is used to extract features from each block and convert them into high-dimensional vectors. Subsequently, these patches are input to a hierarchical Transformer encoder to generate multi-level features with original image resolutions of {1/4, 1/8, 1/16, 1/32}. Efficient self-attention is used in the stage to effectively reduce computational complexity, and a 3×3 convolution operation is introduced to the Hybrid Feedforward Network (Mix-FFN). Mix-FFN can extract global contextual information and preserve local spatial details, significantly enhancing the ability to represent complex structures and subtle features of architectural images. The overlap patch merging technique served as a complementary image processing step, including dividing overlapping blocks, extracting feature, averaging overlapping regions, and merging block to restore overlapping image blocks to a complete image. The decoder abandons the complex attention mechanism or upsampling module in traditional structures and only uses a lightweight multi-layer perceptron (MLP) to fuse the multi-scale features that output by different levels of encoders. The specific process of the full MLP decoder includes four steps: first, the multi-level features that output by the MiT encoder are unified into channel dimensions through the MLP layer; Next, upsample each feature to a resolution of 1/4 of the original image and concatenate them; Subsequently, the concatenated features are fused through the MLP layer; Finally, after another layer of MLP, the segmentation mask with a resolution of H/4×W/4×N cls (N cls is the number of categories) is output. It is worth noting that the overlapping block strategy in SegFormer makes it particularly suitable for handling complex contours of traditional buildings, such as horsehead walls, effectively avoiding the common edge aliasing problem in CNN [ 29 ]. Meanwhile, its powerful self-attention mechanism can establish long-range dependency relationships, accurately associate dispersed building features, and further enhance the accuracy and reliability of traditional building semantic segmentation. Grad-CAM Method Grad-CAM method is used to visualize the classification and segmentation networks [ 30 ]. It can highlight the importance of each position in the feature map of a given category by generating heatmaps. First, the last convolutional layer in the feature extraction module is selected, and the gradient of the target category with respect to feature maps of this layer is calculated. Then, the global average of the gradients on each channel is computed for the k th feature map to obtain the weight of the feature map. Finally, the feature map A k is accumulated according to the weight, and the ReLU function is used to filter out negative values, and the heatmap of category c is obtained. The formula is computed as follows: $$\alpha _{k}^{c}=\frac{1}{Z}\sum\limits_{i} {\sum\limits_{j} {\frac{{\partial {y^c}}}{{\partial A_{{ij}}^{k}}}} }$$ 1 $$L_{{Grad-CAM{\text{ }}}}^{c}=ReLU(\sum\limits_{k} {\alpha _{k}^{c}} {A^k})$$ 2 In the formula, Ak ij represents the pixel value at row i and column j in the k th feature map, Z is the number of pixels in the feature map, y c is the category score of the c th class; αc k is the weight of the k th feature map corresponding to the c th class. A k is the k th feature map of the convolutional layer, and Lc Grad-CAM is the heatmap result of category c . T-SNE Clustering Method T-SNE method is employed to analyze the clustering results of semantic segmentation which is produced based on the OAL images. It is a dimensionality reduction method that reduces the high-dimensional feature of each sample to a low-dimensional space to visualize feature changes [ 31 ]. Samples with similar features are aggregated, while those with different features are separated. The formula is computed as follows: $${p_{j|i}}=\frac{{\exp ( - ||{x_i} - {x_j}|{|^2}/2{\sigma _i}^{2})}}{{\sum\limits_{{}} {_{{k \ne i}}\exp } ( - ||{x_i} - {x_k}|{|^2}/2{\sigma _i}^{2})}}$$ 3 $${p_{ij}}=\frac{{{p_{j|i}}+{p_{i|j}}}}{{2n}}$$ 4 In the formula, x i and x j represent two high-dimensional data points. \(\:\sigma\:\) i is Gaussian kernel width of x i , which is a parameter used to adjust local range of data points, p i|j and p j|i represent the conditional probability of similarity of data points in high-dimensional space, p ij is the joint probability, and n is the number of data points. k is a summation index that iterates over all data points x k except for x i. $${q_{ij}}=\frac{{{{(1+||{y_i} - {y_j}|{|^2})}^{ - 1}}}}{{\sum\nolimits_{{k \ne l}} {{{(1+||{y_k} - {y_l}|{|^2})}^{ - 1}}} }}$$ 5 In the formula, y i and y j are the mapped low dimensional data, and q ij is the conditional probability that the data is similar in the low dimensional space. k and l are summation indices that iterate over all pairs of distinct low-dimensional points y k and y l . $$C=KL(P||Q)=\sum\limits_{i} {\sum\limits_{j} {{p_{ij}}} } \log \frac{{{p_{ij}}}}{{{q_{ij}}}}$$ 6 In the formula, P and Q represent the similarity distribution between data points in high-dimensional and low dimensional space, respectively. KL is Kullback Leibler divergence, a tool used to measure the difference between two distributions. C represents the degree of mismatch between p ij and q ij . Results and Analysis Experimental Preparation The RS and OAL images in the traditional architecture dataset were processed as follows: All images were cropped to 256×256 pixels and divided into training, validation, and test sets at a ratio of 6:2:2. At the same time, data augmentation were used to expand the dataset, such as rotation, mirroring and color enhancement. To ensure the fairness and accuracy of the experiment, the same experimental environment and initial training parameters were used, which is shown in Table 3 . Table 3 Overview of experimental environment and parameters Experimental environment CPU Intel(R) Xeon(R) CPU E5-2673 v4 GPU NVIDIA GeForce RTX 3060 Ti, 12 GB memory CUDA 11.8 PyTorch 2.0.0 Python 3.8 Parameters of double-branch classification network Resize Size= (256,256) Batchsize 32 Epoch 100 Learning rate 1e-2 Optimizer SGD Loss function CrossEntropyLoss Parameters of semantic segmentation network Resize Size= (256,256) Batchsize 8 Epoch 100 Learning rate 1e-4 Optimizer AdamW Weight decay 1e-4 Model Evaluation Metrics To evaluate the performance of the architectural style classification experiments and architectural landscape segmentation experiments, some metrics are introduced here. In the classification experiments, accuracy, precision, recall, and F1-score were employed to evaluate the performance of the double-branch classification network. They are computed as follows: $$Accuracy=\frac{{TP}}{{TP+FP+FN+TN}}$$ 7 $$Precision=\frac{{TP}}{{TP+FP}}$$ 8 $$Recall=\frac{{TP}}{{TP+FN}}$$ 9 10 In landscape segmentation experiments, category average pixel accuracy (mPA), average precision (mPrecision), and average intersection to union ratio (mIoU) are used as evaluation metrics to evaluate the performance of SegFormer. They are computed as follows: $$mPA=\frac{1}{k}\sum\nolimits_{{i=1}}^{k} {\frac{{T{P_i}}}{{T{P_i}+F{N_i}}}}$$ 11 $$mPrecision=\frac{1}{k}\sum\nolimits_{{i=1}}^{k} {\frac{{T{P_i}}}{{T{P_i}+F{P_i}}}}$$ 12 $$mIoU=\frac{1}{k}\sum\nolimits_{{i=1}}^{k} {\frac{{T{P_i}}}{{T{P_i}+F{P_i}+F{N_i}}}}$$ 13 In the above formulas, TP , FP , TN and FN represent true positive, false positive, true negative, and false negative respectively. k represents target categories, and i denotes the object index. Analysis of Architectural Environment Characteristics As the material basis for the formation of traditional settlement landscapes, the natural environment directly affects the spatial layout and morphological characteristics of the settlements. Specifically, terrain is the most crucial element, directly influencing the distribution pattern of water and vegetation within the settlements, as well as the formation of soil [ 32 ]. Water is a necessary condition for human survival, which affects the location and layout of settlements. The selection of vegetation in settlement landscapes is not only based on natural factors, but also on the factors of production and daily life [ 33 ]. Table 4 Statistics of the number of settlement landform types Architectural style Number of settlements Plains Hills Mountains Plateaus Su 40 36 4 Hui 40 4 27 9 Jing 40 3 17 19 1 Min 40 1 7 32 Chuan 40 1 6 33 Jin 40 1 21 18 To have a more comprehensive understanding of the environmental characteristics of architectures for each architectural style, the natural environment, including terrain, water and vegetation, was analyzed using RS and DEM data. As shown in Table 4 , the types of terrain are divided into plain, hill, mountain, and plateau. Specifically, (1) the architectures of Su are mainly distributed in plain areas, which have flat terrain, and the settlements are distributed in high-density blocks. The settlements in this area are either located within river networks or distributed along riverbanks. Artificial vegetation and herbs along the rivers are very common in this area. The garden style layout of the architectures shows the beauty of the integration of architectures and natural landscape. (2) The architectures of Hui are mostly distributed in hills with flat terrain, which is easy to carry out construction activities and can avoid the risk of floods. These settlements are near rivers, and most of the rivers are mountain streams and small rivers. The vegetation coverage in this area is relatively high, and mostly a mixture of arbors and shrubs, forming a continuous landscape with the surrounding forest land. (3) The architectures of Jing are mainly located in hills and mountains. The hilly area has a gentle terrain, and the water is mostly seasonal rivers and artificially excavated ditches. The vegetation is mainly composed of arbor. In the mountainous areas, the height of the mountain is relatively low. The layout of the architectures is adjusted according to the slope and direction. The forest coverage rate is higher in mountainous areas. (4) The architectures of Min are mostly located in wooded mountains or river valleys surrounded by mountains. There are many perennial rivers in the river valley area, and the vegetation in the mountains is mostly broad-leaved forests. The rivers and closed terrain together form a natural barrier for family settlement. (5) The architectures of Chuan are mostly located in mountainous areas with steep terrain. The cantilevered structure of the stilt house adapts to mountain slopes, which is convenient for ventilation. There are numerous rivers, and the surrounding vegetation is mainly bamboo forests and broad-leaved shrubs. (6) The architectures of Jin are mainly distributed in mountains and plateaus, and the terrain mainly consists of steep mountains and the Loess Plateau. The buildings in mountainous areas adopt deep eaves and thick walls, which can withstand severe cold and temperature changes. The water is mainly mountain streams and seasonal rivers, and vegetation is scattered shrubs with a relatively low coverage. The terrain of the Loess Plateau is undulating, and architectures here make full use of loess resources. The cave dwellings are typical architectures. The water is dominated by seasonal rivers, and the vegetation coverage is low. Interpretation of Architectural Features In this chapter, the architectural styles are firstly classified to achieve intelligent recognition of architectural styles. Then, using Grad-CAM, representative features of each architectural style are obtained. Finally, architectural features are interpreted for different architectural styles. Classification of Architectural Styles In order to select the most suitable network to class the architectural style, ResNet152, MobileNetV2, EfficientNet-B0 were used to replace the network in the feature extraction modules for comparative experiments, and these models were named RDB-Net, MDB-Net, EDB-Net, respectively [ 34 – 36 ]. In the experiment, all models used the same training and testing datasets for fairness. Table 5 lists the evaluation metrics and their average precision values of the comparative models. The results show that the proposed model achieved the highest average value of three indicators, namely, recall (89.73%), accuracy (90.37%), and F1-score (89.60%). The proposed model has been identified as the most effective model for identifying architectural styles. Table 5 Evaluation indicators for classification models Style Su Hui Jing Min Chuan Jin Average MDB-Net Recall 89.55 94.03 68.66 89.55 96.97 81.82 86.76 Precision 88.24 91.30 76.67 92.31 92.75 78.26 86.59 F1-score 88.89 92.65 72.44 90.91 94.81 80.00 86.62 RDB-Net Recall 85.07 88.06 71.64 91.04 96.97 75.76 84.76 Precision 82.61 76.62 82.76 81.33 91.43 98.04 85.47 F1-score 83.82 81.94 76.80 85.92 94.12 85.47 84.68 EDB-Net Recall 80.60 95.52 79.10 88.06 93.94 74.24 85.24 Precision 93.10 82.05 68.83 96.72 96.88 79.03 86.10 F1-score 86.40 88.28 73.61 92.19 95.38 76.56 85.40 Ours Recall 92.54 98.51 88.06 89.55 98.48 71.21 89.73 Precision 93.94 89.19 77.63 95.24 90.28 95.92 90.37 F1-score 93.23 93.62 82.52 92.31 94.20 81.74 89.60 Based on the classification model, the highlight areas in the Grad-CAM heatmap of OAL images can visualize representative features of the architectural style, as shown in Fig. 6 . The brighter areas in the heatmap represent the regions that the model focuses on, and the redder the color, the higher the attention it receives. The model focuses on the architecture itself, rather than other backgrounds, which verifies the reliability of the classification model. Interpretation of Architectural Features for Each Architectural Styles. According to Fig. 6 , the model focuses on both the overall structure and the detailed elements of the architectures. To explore the key characteristics that affect the classification of architectural style, architectural components, roof types, architectural colors and architectural materials are analyzed in this chapter. The features with the frequency of less than 10 times are not considered as the representative features, and thus it will not be discussed here. 1. Architectural components Figure 7 presents the types and definitions of architectural components. According to the Grad-CAM heatmaps, the attention frequency of different architectural components for each architectural style is counted, which is shown in Fig. 8 . As the important component of architectures, the roof has the highest attention frequency in all architectural styles. It reflects the crucial role of the roof in shaping architectural styles. The roofs of different architectural styles have diverse forms and characteristics, which will be discussed in the next subsection. Except for the roof, wooden windows are the main components in all architectural styles. The horsehead wall is the key component of Hui. Plaques often appear in the Jin and Min, while stone windows often appear in the Su and Hui. Specifically, (1) in the architectures of Su, wooden windows are the components that receive the most attention except for the roof. These windows emphasize decoration, and integrate with surrounding landscapes, such as mountains, water, flowers and trees. As a leaky window in garden architectures, the rich patterns of stone windows make them highly decorative. Columns and railings, as essential elements of garden architectures, pay attention to carving and decoration, and have a high frequency of attention. (2) In the architectures of Hui, horsehead walls are the most recognizable components except for the roof. Their structures can prevent fires. Doors with exquisite brick and wood carvings are important symbols of art and family wealth. (3) In the architectures of Jing, doors are the components that receive the most attention except for the roof. Their shape and scale adhere strict hierarchy and regulations, representing the family status and wealth level. The attention frequency of columns is also relatively high, and their arrangement and size follow strict construction regulations. (4) In the architectures of Min, wooden windows have the most attention except for the roof. The wooden windows have white frames and simple forms of smaller size. The gates of the Tulou are heavy and sturdy, equipped with solid locks and defense facilities. Plaques often display the family name, or auspicious words. (5) In the architectures of Chuan, railings are the components that receive the most attention except for the roof. Numerous railings have been set up in platforms, pavilions and other spaces. The arrangement and size of columns are designed according to the terrain characteristics. The overall height of the structures can be adjusted to adapt to complex terrain. (6) In the architectures of Jin, doors have the most attention except for the roof, and their shape and scale are extremely exquisite, decorating with brick carvings. Plaques are commonly found in temples and ancestral halls. The decoration consists of lanterns and stone carvings. Lanterns are commonly used in festival scenes, while stone carvings are applied to lintels and column foundations. 2. Roof type The types and characteristics of the roofs are shown in Table 6 . The Grad-CAM heatmaps were used to calculate the attention frequency of different roof types for each architectural style, as shown in Fig. 9 . Overall, hard mountain roofs are widely used in the architectures of Su, Hui, Jing, and Jin. Overhanging gable roofs are used in the architectures of Min and Chuan, and pyramid roofs are used in Su and Chuan. Specifically, (1) in the architectures of Su, the gable-on-hip roofs are primary roof styles, which are commonly used in ancestral halls and temples. The pyramid roofs frequently appear in Jiangnan gardens paired with pavilions and towers, which serve as prominent visual focal point in the landscape. Hard mountain roofs are commonly used in residential buildings to meet the practical needs of fire and wind prevention. (2) In the architectures of Hui, hard mountain roofs are the primary roof types. The gable-on-hip roofs are primarily used in ancestral halls, temples, or similar structures to emphasize their functional importance. (3) In the architectures of Jing, hard mountain roofs are commonly used in residential buildings, characterized by low construction cost. The hip roofs are main roof types of the palaces, symbolizing the supreme imperial power. In addition to hip roofs, gable-on-hip roofs are also found on the palace side hall, temples and some official buildings, serving both aesthetic and hierarchical purposes. (4) In the architectures of Min, overhanging gable roofs are the main roof type. The large overhanging eaves of Tulou not only provide shade for walls and doors but also facilitate effective drainage. (5) The overhanging gable roofs are primary roof types in the architectures of Chuan, which play a role in drainage. Pyramid roofs and gable-on-hip roofs are main roof types for bamboo houses, drum towers, and wind-rain bridges, meeting needs for sightseeing, relaxation, and clan activities. (6) Hard mountain roofs are commonly used in the architectures of Jin, and they can effectively resist the erosion of wind and sand on the building walls and roofs. Flat roofs are common in cave buildings, and they matche the structure of the cave, providing the space for drying crops and storing debris. 3. Architectural colors According to OAL images, architectural colors of each image were counted, as shown in Table 7 . It can be seen from the table that main colors of the architectures of Su and Hui are black, white and gray. Golden and vermillion are the common color in the palaces of Jing, while black and reddish brown are main color in the residential buildings. Specifically, (1) in the architectures of Su, walls are off white, and roofs are covered with small black tiles. The door and window columns have two colors, one is black or dark brown, and the other is reddish brown. (2) The architecture of Hui is deeply influenced by Confucian culture, and most of the columns of doors and windows are black or dark brown. (3) In the architectures of Jing, palaces predominantly feature vermilion walls and golden glazed tiles, and the columns of doors and windows are also vermilion. The color conforms to the aesthetic pursuit of solemnity and magnificence in royal culture. The walls of residential buildings are white or tan. Door and window columns are primarily reddish brown or dark brown, with reddish brown being more common. (4) In the architectures of Min, the exterior walls are rammed with local soil, resulting in an earth yellow color. The doors are commonly used reddish brown, which has the meaning of exorcising evil in Hakka culture. (5) In the architectures of Chuan, the regions where the architectures are rich in bamboo and wood resources, and the main color of the architectures is dark brown. (6) In the architectures of Jin, the wall color of Shanxi merchants' architecture is dark gray, with a small amount of earth yellow. The wall color of the cave is mainly earth yellow. Table 7 Quantity statistics of architectural colors based on OAL images Style Type Wall color Color of door and window columns Roof color Count Su Off white Dark brown Black 205 Off white Reddish brown Black 129 Hui Off white Dark brown Black 309 Off white Reddish brown Black 25 Jing Palace Vermilion Vermilion Golden 98 Quadrangle White or tan Reddish brown Black 157 White or tan Dark brown Black 79 Min Tulou Earth yellow Reddish brown Black 329 Off white and reddish brown Reddish brown Black 5 Chuan Stilted building Dark brown Dark brown Dark gray 270 Drum tower and Wind-rain bridge Dark brown Dark brown and white Dark brown 55 Bamboo building Dark brown Dark brown Dark brown 9 Jin Shanxi Merchants' Architecture Dark gray Dark brown and reddish brown Dark gray 164 Earth yellow Dark brown Dark gray 20 Cave dwelling Dark gray Dark brown and reddish brown Dark gray 23 Earth yellow Dark brown Dark brown 127 4. Architectural materials Based on OAL images, architectural materials used for each style were statistically analyzed, as shown in Table 8 . Southern architectures primarily use bamboo, wood, and rammed earth. Local materials are used for humid climates and complex topography. Northern architectures utilize bricks, stones, tiles, and loess to adapt to arid conditions. Table 8 Statistics of architectural materials based on OAL images Style Type Wall material Roof material The material of door and window columns Count Su Stone and brick Tiles Wood 334 Hui Stone and brick Tiles Wood 334 Jing Palace Stone and brick Glazed tiles Wood and white marble 98 Quadrangle Stone and brick Tiles Wood 236 Min Tulou Rammed earth Tiles Wood 334 Chuan Stilted building Wood Tiles or thatch Wood 270 Drum tower Wood Tiles Wood or earth-stone 14 Wind-rain bridge Wood Tiles Wood or stone 41 Bamboo building Bamboo-wood mixture Thatch Bamboo-wood mixture 9 Jin Shanxi Merchants' Architecture Stone and brick Tiles Wood 184 Cave dwelling Loess Wood and thatch Wood 110 Loess, stone and brick Tiles, stone and brick Wood 40 Specifically, (1) the architectures of Su and Hui are consistent in materials. The walls are made of solid bricks, the roofs are made of small blue tiles, which play a rainproof role. The columns of door and window are made of wood, and they are good at making wood carvings for decoration. (2) In the architectures of Jing, palace roofs are covered with glazed tiles, and door and window columns are made of white marble stone, highlighting the elegance and magnificence. The roofs of residential buildings are paved with ordinary tiles. (3) In the architectures of Min, the wall materials take rammed earth as the core materials, and its components are clay, sand and lime, mixed with bamboo to enhance the structural strength. It has the functions of heat preservation, heat insulation and defense. (4) In the architectures of Chuan, stilted buildings use wood as the material of the columns of walls and door columns, and the roof uses small green tiles or straw. The roofs of bamboo tower of Dai are covered with thatch and made of a mixture of bamboo and wood materials, which can resist the erosion of rainwater and insects. (5) In the architectures of Jin, the walls of Shanxi merchants' architectures are made of bricks, offering high defensive capability and effective resistance to cold and sandstorm. The roofs are covered with tiles to accelerate drainage. Small sized cave dwellings use loess as the wall material to achieve the purpose of warm in winter and cool in summer. In the larger sized courtyards, the roofs are made of bricks and tiles with stronger load-bearing capacity, and the walls are made of a mixed structure of loess and bricks, which enhances durability through material composites. Interpretation of Architectural Landscape Features To find the most representative architectural landscape features for each architectural style, the semantic segmentation experiment in OAL images is firstly carried out. Combined with the Grad-CAM heatmaps, the representative architectural landscape elements are selected. Then, the semantic clustering results are analyzed with T-SNE model to explore the semantic association and potential confusion reasons for each architectural style. Semantic Segmentation of Architectural Landscape To select the most suitable model to segment the architectural landscape elements, several common segmentation models were compared, such as Segmenter, DeepLabV3+, UNet, PSPNet, Swin Transformer and SegFormer. In the experiment, all models used the same training and testing datasets for fairness. Table 9 lists the evaluation metrics and the average precision of the comparative models for each segmentation elements. The results show that SegFormer achieves the highest performance among the average values of three indicators: pixel accuracy (PA), accuracy (P), and Intersection over Union (IoU). Therefore, the SegFormer was selected as the model to segment the architectural landscape elements on OAL images. Table 9 Evaluation indicators of segmentation models Model Segmenter DeepLabV3+ UNet PSPNet Swin Transformer SegFormer mPA 68.41 79.04 80.30 82.13 83.29 84.11 mPrecision 83.87 86.94 87.55 89.35 87.97 89.92 mIoU 61.97 70.49 72.67 74.77 74.82 76.44 Sky PA 96.13 96.73 97.24 96.09 98.24 97.03 P 93.56 95.25 96.78 95.17 96.72 98.14 IoU 90.16 92.28 94.20 91.63 95.07 95.28 Building PA 94.17 94.38 94.13 94.19 94.90 94.99 P 90.91 93.44 93.74 93.25 95.12 95.22 IoU 86.07 88.51 88.56 88.18 90.49 90.67 Road PA 77.73 86.44 88.07 88.39 89.48 90.88 P 90.94 85.27 85.95 85.95 89.68 87.75 IoU 72.15 75.21 76.99 77.23 81.13 80.65 Mountain PA 92.02 93.49 93.77 93.28 91.95 93.86 P 89.30 92.89 90.60 93.11 91.29 93.35 IoU 82.88 87.18 85.46 87.25 84.53 87.98 Vegetation PA 85.78 87.03 87.73 86.68 90.03 90.77 P 86.30 88.93 88.89 88.89 89.04 89.95 IoU 75.50 78.52 79.05 78.21 81.05 82.41 Water PA 89.90 94.01 93.07 94.84 94.53 95.45 P 95.67 92.09 95.14 93.49 95.51 93.30 IoU 86.38 86.98 88.84 88.96 90.51 89.33 Stone steps PA 45.94 61.88 66.52 62.92 58.57 67.59 P 83.59 86.40 82.94 88.40 83.37 88.64 IoU 48.75 56.39 58.52 58.12 52.45 62.21 Stone balustrade PA 72.72 80.05 81.00 83.17 85.62 85.68 P 72.69 82.46 78.85 76.76 88.90 81.96 IoU 57.11 68.40 66.54 66.44 77.35 72.08 Well PA 42.23 88.75 87.60 91.69 94.79 90.61 P 59.57 68.25 91.22 90.18 82.69 91.75 IoU 50.30 62.82 80.79 83.37 79.10 83.78 Bridge PA 63.31 78.18 75.69 77.79 81.42 77.73 P 79.80 81.96 86.68 86.71 82.38 89.50 IoU 54.57 66.70 67.80 69.50 69.34 71.23 Gate pier PA 53.80 62.16 63.41 73.30 69.58 91.02 P 90.61 84.46 81.62 85.27 81.08 91.55 IoU 40.22 55.78 55.49 65.06 59.86 83.97 Vat PA 44.36 88.33 74.87 91.74 97.11 92.87 P 98.21 91.14 76.38 92.53 75.86 72.55 IoU 44.36 81.35 60.79 85.41 74.18 68.73 Stone PA 85.31 79.53 82.03 83.62 88.46 82.32 P 74.24 84.05 85.58 85.42 87.69 85.79 IoU 65.82 69.10 72.07 73.18 78.69 72.44 Censer PA 7.24 17.39 27.69 35.50 26.43 29.12 P 64.58 87.61 82.47 91.78 86.82 95.99 IoU 6.71 16.97 26.15 34.40 25.41 28.77 Stone mill PA 75.58 77.28 91.73 78.76 88.27 81.65 P 88.12 90.05 96.47 93.35 93.40 93.30 IoU 68.60 71.20 88.74 74.58 83.09 77.13 Representative Architectural Landscape of Different Architectural Styles Based on the segmentation results of the architectural landscape elements, the attention frequency of the Grad-CAM heatmaps on each architectural style is counted, as shown in Fig. 10 . The statistical results reflect the significance of different architectural landscape elements in the classification process for each architectural style. These statistics can better reflect the repeatability and stability of the architectural landscape elements in architectural style discrimination. The more frequently it occurs, the greater the contribution of the landscape element. It should be pointed out that elements with the frequency of less than 10 times are not considered as the representative element, and thus they will not be discussed here. For each architectural style, the top six architectural landscape elements with the highest frequency are selected as the representative landscape elements from 14 kinds. Sky, road, and vegetation are landscape elements that all architectural styles pay attention to. Water, mountain, stone, well and bridge commonly appear in the southern architectural landscape; Gate pier, vat, center and stone mill commonly appear in the northern architectural landscape. Stone steps and stone balustrade commonly appear in the architectures of Jing. In addition, each architectural style also has some unique landscape elements, which have unique cultural connotation and landscape value. Specifically, (1) in the architectures of Su, the representative architectural landscape elements are vegetation, sky, water, stone, road and bridge. Water, bridges, and stones are the core landscape symbols of garden architectures. These elements highlight the garden feature by the integration of architectures and nature. (2) In the architectures of Hui, the representative architectural landscape elements are vegetation, sky, mountains, water, roads and bridges. The combination of mountains and water highlights the layout of settlements. The bridge is an important connection, and stone arch bridges and stone slab bridges are relatively common. The gate pier is located on both sides of the gate, carved with exquisite patterns. Stone steps are commonly found at the entrance of important architectures such as ancestral halls and temples, with a simple and solid structure. (3) In the architectures of Jing, the representative landscape elements are sky, roads, vegetation, stone steps, mountains and stone balustrade. Stone steps and stone balustrade are the key landscape elements. Stone balustrade carved with dragons and phoenixes is widely used in the palaces such as the Forbidden City, which highlights the dignity of the royal family. Stone steps are used in palaces and temples, and their materials, widths, and levels are governed by strict regulations for different architectural grades. The door piers located on both sides of the gate are carved with lions and drums, which symbolize social status. The vat with goldfish and lotus is commonly placed in the courtyard, which adds vitality and elegance to the courtyard. Censer is commonly placed in the temples and palaces, and is used for activities of sacrifice. (4) The representative landscape elements of Min are mountains, vegetation, sky, roads, water and wells. Wells are the key landscape element with regional characteristics. They have met the residents' demand for domestic water under the settlement pattern. (5) The representative landscape elements of Chuan are mountains, sky, vegetation, water, roads and stone balustrade. As the protective and decorative element, stone balustrade is commonly used at the edge of bridges, plank roads and building platforms. Their carving patterns are integrated with the folk culture elements of the Bashu region. (6) In the architectures of Jin, the representative landscape elements are the sky, roads, mountains, vegetation, stone steps and gate piers. The gate piers are often carved with rich patterns. Together with the stone steps, they highlight the architectural rituals and defense needs in the Loess Plateau. The vat not only can store water, but also can adjust the microclimate in the courtyard. The stone mill is a symbol of traditional agricultural culture, and it was once an indispensable production tool in rural life. Semantic Clustering Analysis The semantic clustering analysis is carried out with T-SNE model to explore the semantic association and potential confusion reasons of representative architectural landscape elements for each architectural style. It can further understand the cognitive logic of the model to traditional architectural landscapes. Figure 11 shows the semantic clustering results of representative architectural landscapes for each architectural style. Specifically, (1) in the architectures of Su, the semantic clustering of the landscapes presents distinctive garden features. As the unique semantic feature of the Su, vegetation, stone and water together create a natural landscape, which leads to semantic confusion. (2) In the architectures of Hui, laying roads along water is a typical layout of Hui, and the spatial dependence leads to a blurred semantic boundary between them. In addition, the mountains are covered with vegetation, leading to semantic confusion. (3) In the architectures of Jing, the unique semantic connection of landscape elements is reflected in the association of stone balustrade, stone steps and roads. Stone steps are used as the extension of roads, and stone balustrade is defined as the scope of roads. Semantic confusion occurs due to the closeness of their spatial positions. (4) In the architectures of Min, as important life facilities in Tulou buildings, wells are directly connected to internal roads, which leads to semantic confusion. In addition, the mountains are covered with vegetation, leading to semantic connection. (5) In the architectures of Chuan, mountains, vegetation, and water form a close semantic connection. In addition, vegetation around the architectures and the stone railings around the stilt houses intersect in space, and the relationship between architectures and nature creates a semantic confusion. (6) For the architectures of Jin, the semantic association of landscape elements highlights the environmental characteristics of the Loess Plateau. The stone steps are located at the entrance of the building, and the gate piers are adjacent to the stone steps and stand on both sides of the road. The tight structure of them leads to semantic confusion. Discussions The innovation points in this paper are summarized as follows: (1) multi-source data fusion and dedicated dataset construction. To break the limitations of single data sources, RS, DEM, and OAL images are integrated to identify and interpret the style characteristics of Chinese traditional architectural heritage. According to the list of CTSs, the CTS directory of architectural styles is established, and typical samples are selected and labeled. A set of special datasets suitable for interpretation of style characteristics of traditional architectures is constructed. (2) The perspective of Chinese traditional architectural style. The existing research mostly focuses on the macro settlement pattern or single architectural type, and neglects the landscape features of the architectural styles. With the help of DL technology, this paper quantitatively revealed the representative characteristics and characteristic combination rules for each architectural style, such as environmental characteristics, architectural features and architectural landscape characteristics. (3) Construction of hierarchical analysis framework of “architectural environment characteristics - architectural features- architectural landscape characteristics”. Natural environment for each architectural style is firstly analyzed, and then architectural features for each style, which include architectural components, roof types, architectural colors and materials, are analyzed by using the classification network and visualization technology. Finally, the representative architectural landscape elements are identified by using semantic segmentation model. This framework elucidates the inherent coupling relationship between natural environment, architecture, and landscape semantics, providing a systematic theoretical basis for the research and protection of traditional architectural heritage. Based on Grad-CAM visualization analysis, we found that the model not only focuses on the architectural features, but also relies on the surrounding architectural landscape elements, when discriminating architectural styles. Both of them reflect the collaborative shaping of regional environment, cultural influence and functional requirements on traditional architectures. (1) Regional environment. In terms of roof type, the architectures of Min and Chuan have different roof shapes, but both utilize overhanging gable roofs. The southern region, where these architectures are located, has a rainy climate, and the structure of overhanging gable roofs can reduce the erosion of rainwater on the architectures. For the architectural materials, the southern regions are rich in forests, and the architectures of the Su, Hui and Chuan mostly use wood as architectural material. The winter in northern China is cold, dry and windy, and the architectures of Jing and Jin widely use masonry and rammed earth as architectural materials to prevent wind and keep warm. (2) Cultural influence. The architectures of Su construct the natural landscape artistic conception through landscape elements such as water, bridge and stone. The architectures of Hui are deeply influenced by the Confucian culture, and it emphasizes simplicity and solemnity in color. In the architectures of Jing, not only does the shape and decoration of the door strictly comply with etiquette norms, but also stone steps, stone balustrade and gate piers also strengthen the grade through materials and shapes. The architectures of Jin are deeply influenced by farming culture. Vat and stone mill not only meet daily practical needs, but also serve as the material carrier of farming production, which reflect regional economic form and cultural tradition. (3) Functional requirements. In terms of wind and fire prevention, the horsehead walls of the architectures of Hui and its hard mountain roofs highlight the function. For the drainage, both the architectures of Min and Chuan widely use the overhanging gable roofs to achieve effective drainage. For the water security, as the key landscape elements, wells meet the demand of domestic water. It should be noted that some important landscape elements, such as stone carvings, wells, stone mills, are not paid much attention to during the identification. It may be related to insufficient training samples or low element dominance during image acquisition, but they are still the key elements to the identification of architectural style, which carry unique cultural connotation. There are still some limitations in our study, which include two aspects: on the one hand, the scale of dataset is smaller, and a small number of images downloaded from the network belong to the wrong architectural styles. It may have a negative impact on the analysis of characteristics of architectural styles. On the other hand, the current model is better at identifying visually significant and structural features, such as roofs and horsehead walls, and not good at the small size of landscape elements that rely on other semantic context, such as censers. To address these limitations, future work can focus on the following aspects: (1) Expanding the high-quality image dataset by incorporating images from professional map platforms, academic databases, and live photos taken by residents. This approach will improve the accuracy and comprehensiveness of the data. (2) Developing more advanced algorithms that integrate geographic location and cultural prior knowledge to enhance the model's understanding of architectural style. (3) Expanding the research scope to include historical images, which can reveal the spatial evolution patterns and internal mechanisms of traditional architectural heritage and settlements. Conclusion Based on RS, DEM and OAL images, style characteristics of Chinese traditional architectures are identified and interpreted using deep learning and Grad-CAM technology. The hierarchical analysis framework of “architectural environment characteristics - architectural features- architectural landscape characteristics” is constructed, which reveals the architectural characteristics from the perspective of architectural style. Natural environment for each architectural style is firstly analyzed, and then architectural features for each style, which include architectural components, roof types, architectural colors and materials, are analyzed by using the classification network and visualization technology. Finally, the representative architectural landscape elements were identified by using semantic segmentation model. It explains the intrinsic coupling relationship among natural environment, architectures and landscape semantic expression. This research proves that deep learning is an effective method to interpret style characteristics of traditional architectures, offering a novel pathway for the digital conservation and stylistic studies of traditional architectural heritage. In the future, high-quality images, advanced algorithms, and other research on traditional architectures can be further expanded. Declarations Funding This work was supported in part by the Science Foundation of Hebei Normal University under Grant L2025K04, and in part by the Science and Technology Project of Hebei Education Department under Grant BJK2022031. Author Contribution C. C.Y., L.S.D., D.J.Y., H.Y.C., and F.Z.Y. collected data, conducted experimental work, and analyzed and verified the results. C.C.Y. and L.S.D. developed the methodology and wrote the manuscript. All authors reviewed the manuscript. Data Availability Appendix A is provided within the supplementary information files. References Zeng C, Liu PL, Huang LQ. Measuring the fragmentation of traditional village architectural landscape from the perspective of landscape diversity and heterogeneity. Sci Geogr Sin. 2023;43:1973–82. Wu LY. Zhongguo ren ju shi [History of Chinese Human Settlements]. China Architecture & Building; 2014. pp. 215–28. Liu PL. On Construction and Utilization of Chinese Traditional Settlements Landscape's Genetic Map. Peking University; 2011. Hu Z, Liu PL, Deng YY, Zheng WW. Research on identification and extraction methods of traditional settlement landscape genes. Sci Geogr Sin. 2015;35:1518–24. Duan YP, Chen MX, Liu Y, Wang Y, Zhang L. Research on the Cultural Landscape Features and Regional Variations of Traditional Villages and Dwellings in Multicultural Blending Areas: A Case Study of the Jiangxi-Anhui Junction Region. Appl Sci. 2025;15:2185. Ren P, Hong B, Liu Y, Zhou J. Spatial change characteristics of rural settlements and their impacts on landscape patterns based on RS and GIS. Acta Ecol Sin. 2014;34:3331–40. Wang JW. The syntactic graph interpretation of the traditional settlement space from the perspective of cognitive schema: A study of four villages of Han nationality and Zhuang nationality in northern Guangxi. J Beijing Univ Soc Sci. 2024;23:109–17. Li L. Establishment of characteristic atlas of traditional settlements landscape in multicultural area: Take southern Shaanxi for example. Areal Res Dev. 2020;39:175–80. Xing Y, Chen HL, Xing YM. A model for landscape heritage information extraction and 3D generation from sketches and text prompts with IoT-enhanced data fusion. Alexandria Eng J. 2025;129:937–45. Lu Y, LSEVGG. An attention mechanism and lightweight-improved VGG network for remote sensing landscape image classification. Alexandria Eng J. 2025;127:943–51. Mathias M, Martinovic A, Weissenberg J, Haegler S. Automatic architectural style recognition. ISPRS Int Arch Photogramm Remote Sens Spat Inf Sci. 2012;XXXVIII–5/W16:1–6. Han Q, Yin C, Deng YY, Liu PL. Towards classification of architectural styles of Chinese traditional settlements using deep learning: A dataset, a new framework, and its interpretability. Remote Sens. 2022;14:5250. Tan GY, Zhu JK, Chen ZX. Deep learning based identification and interpretability research of traditional village heritage value elements: A case study in Hubei Province. Herit Sci. 2024;12:200. Dai ML, Ward WOC, Meyers G, Tingley DD, Mayfield M. Residential building facade segmentation in the urban environment. Build Environ. 2021;199:107921. Haznedar B, Bayraktar R, Ozturk AE, Arayici Y. Implementing PointNet for point cloud segmentation in the heritage context. Herit Sci. 2023;11:2. Sun RF, Zhang SL, Su MH, Hu LH, Zhang JF. Weakly supervised semantic segmentation for ancient architecture based on multiscale adaptive fusion and spectral clustering. Int J Appl Earth Obs Geoinf. 2025;126:104164. Xiong YZ et al. IEEE,. Accurate detection of historical buildings using aerial photographs and deep transfer learning. In IGARSS 2020–2020 IEEE International Geoscience and Remote Sensing Symposium 1592–1595 (2020). Du LY, Wang YS, Bi. -YOLO: A novel object detection network and dataset for components of China heritage buildings. J Build Eng. 2024;97:110817. Obeso AM, Vazquez MSG, Acosta AAR, Benois-Pineau J, Connoisseur. Classification of styles of Mexican architectural heritage with deep learning and visual attention prediction. In Proceedings of the 15th International Workshop on Content-Based Multimedia Indexing 1–7ACM, (2017). Sun ML, Zhang F, Duarte F, Ratti C. Understanding architecture age and style through deep learning. Cities. 2022;128:103787. Zu XY, Gao C, Wang Y. Interpreting regional characteristics of Tibetan-Qiang houses in Northwestern Sichuan by deep learning and image landscape. Int J Appl Earth Obs Geoinf. 2024;129:103865. Pan ZK, Xu JS, Guo YB, Hu YM, Wang G. X. Deep learning segmentation and classification for urban village using a worldview satellite image based on U-Net. Remote Sens. 2020;12:1574. Chen QT, Zhang L, Duan YP, Fan SQ. Spatial-temporal pattern and evolution of traditional villages in Jiangxi province. Natl Remote Sens Bull. 2021;25:2460–71. Huang G, Liu Z, Van Der Maaten L, Weinberger K. Q. Densely connected convolutional networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2261–2269IEEE, (2017). Woo S, Park J, Lee JY, Kweon IS. CBAM: Convolutional block attention module. In Computer Vision – ECCV 2018. Springer; 2018. pp. 3–19. Zhang XX, Li SD, Chen CY. Classification of architectural styles in Chinese traditional settlements using remote sensing images and building facade pictures. J Geogr Sci. 2024;34:2457–76. Xie EZ et al. Curran Associates,. SegFormer: Simple and efficient design for semantic segmentation with transformers. In Advances in Neural Information Processing Systems Vol . 34 12077–12090 (2021). Zheng SX et al. IEEE,. Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 6877–6886 (2021). Wang WH et al. IEEE,. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions. In Proceedings of the IEEE/CVF International Conference on Computer Vision 548–558 (2021). Selvaraju RR et al. IEEE,. Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE International Conference on Computer Vision 618–626 (2017). Hinton GE, Roweis S. Stochastic neighbor embedding. Adv Neural Inf Process Syst. 2002;15:833–40. Liu LM. History and Perspective of Development of Rural Landscape Planning in China. J Ecol Rural Environ. 2001;17:52–5. Pei BJ. The Study about Natural and Geographical Factors on Rural Landscape Spatial Structure in Northeastern Hainan Island. Hainan University; 2015. He KM, Zhang XY, Ren SQ, Sun J. IEEE,. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 770–778 (2016). Sandler M, Howard A, Zhu ML, Zhmoginov A, Chen LJ. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition 4510–4520IEEE, (2018). Tan MX, Le QV, Efficientnet. Rethinking model scaling for convolutional neural networks. In Proceedings of the 36th International Conference on Machine Learning 6105–6114PMLR, (2019). Tables 1 and 6 Tables 1 and 6 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files AppendixA.docx Table1and6.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 25 Nov, 2025 Reviews received at journal 25 Nov, 2025 Reviews received at journal 24 Nov, 2025 Reviews received at journal 19 Nov, 2025 Reviews received at journal 17 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviews received at journal 06 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviewers agreed at journal 06 Nov, 2025 Reviewers agreed at journal 05 Nov, 2025 Reviewers agreed at journal 05 Nov, 2025 Reviewers agreed at journal 05 Nov, 2025 Reviewers agreed at journal 05 Nov, 2025 Reviewers invited by journal 05 Nov, 2025 Editor assigned by journal 31 Oct, 2025 Submission checks completed at journal 31 Oct, 2025 First submitted to journal 28 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7971684","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":541296415,"identity":"c0826d16-9adc-4f5c-ae09-942980123d7a","order_by":0,"name":"Changyao Chen","email":"","orcid":"","institution":"Hebei Normal University","correspondingAuthor":false,"prefix":"","firstName":"Changyao","middleName":"","lastName":"Chen","suffix":""},{"id":541296416,"identity":"a364f726-aec2-4344-a428-fbb4e4d73491","order_by":1,"name":"Shaodan Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIiWNgGAWjYBACPmYGNiDFzMDA3gMiQSABvxY2uBaeM0AygRgtDDAtEjnEamFnf/bgR4W1vLnk24OPC38cZuBnzzFg+LkDn8N4zA17zqQb7pydl2w8I+Ewg2TPGwPG3jN4tbBJ8LYdZtxwO8dMmgeoxeBGjgEzYxs+LezPJP+2HbbfcPOM+W+QFnvCWhjMpIG2JG64wWPGDLZFgqAWHjNpmTPpyRvO5BhLz0hL55E486zgYC8eLfz8x59Jvqmwtt1w/Izh5wIbazn+9uSND37i0YIBeEDEARI0jIJRMApGwSjAAgDRwkfC4q2oKgAAAABJRU5ErkJggg==","orcid":"","institution":"Hebei Normal University","correspondingAuthor":true,"prefix":"","firstName":"Shaodan","middleName":"","lastName":"Li","suffix":""},{"id":541296417,"identity":"a097654a-cb3e-4598-b18f-9d9d683ab572","order_by":2,"name":"Jiayou Ding","email":"","orcid":"","institution":"Hebei Normal University","correspondingAuthor":false,"prefix":"","firstName":"Jiayou","middleName":"","lastName":"Ding","suffix":""},{"id":541296418,"identity":"01217cae-c5d6-461f-8035-598dd21c703a","order_by":3,"name":"Yincui Hu","email":"","orcid":"","institution":"Hebei Normal University","correspondingAuthor":false,"prefix":"","firstName":"Yincui","middleName":"","lastName":"Hu","suffix":""},{"id":541296419,"identity":"b08490d6-57c7-4ce2-871a-40ffe26323fa","order_by":4,"name":"Ziyue Fan","email":"","orcid":"","institution":"Hebei Normal University","correspondingAuthor":false,"prefix":"","firstName":"Ziyue","middleName":"","lastName":"Fan","suffix":""}],"badges":[],"createdAt":"2025-10-28 17:18:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7971684/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7971684/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":96050030,"identity":"ba501bd1-9bcc-4ff9-bcf4-46243c55b423","added_by":"auto","created_at":"2025-11-17 06:36:07","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4904173,"visible":true,"origin":"","legend":"","description":"","filename":"IdentifyingandInterpretingTraditionalArchitecturalStyleCharacteristicsBasedonDeepLearning.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cc81ea3d7079555fae18beed.docx"},{"id":96050020,"identity":"182abfde-4fe0-43f0-a4e0-53cd431edbae","added_by":"auto","created_at":"2025-11-17 06:36:05","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":901896,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f2dc72f6f308b216aaa4ff9c.docx"},{"id":96247494,"identity":"a6f367b3-e9dd-4a41-86cf-252f13af5b0a","added_by":"auto","created_at":"2025-11-19 07:27:31","extension":"jpg","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":529247,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/2f27508bd86dd9e9b01b63df.jpg"},{"id":96050032,"identity":"a0a5ff85-8a43-4cfa-87dd-5b23a4e4194c","added_by":"auto","created_at":"2025-11-17 06:36:07","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13130,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7c5d1f6775f7964f9e96d325.docx"},{"id":96050043,"identity":"3d45f32f-63ce-4e9f-a05b-12690804451a","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69907,"visible":true,"origin":"","legend":"","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5cf0dcdcb016f2e5b6739458.png"},{"id":96050023,"identity":"01e1e8c9-43f1-4f10-9083-1384a6bdc149","added_by":"auto","created_at":"2025-11-17 06:36:06","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13820,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cad30b8c809743ce5c6291b2.docx"},{"id":96050059,"identity":"517a38c7-c9e0-49cc-92bd-45477ccc987a","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4924685,"visible":true,"origin":"","legend":"","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0466b9ce3ca8cb46941e4452.png"},{"id":96050037,"identity":"43faf294-8175-4965-be09-92b717b05191","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13533,"visible":true,"origin":"","legend":"","description":"","filename":"Table4.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7f13eb9e669284c44024147b.docx"},{"id":96246657,"identity":"5ffa2b92-a26c-4050-a0a2-1f3f5bb47558","added_by":"auto","created_at":"2025-11-19 07:26:27","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3547157,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1384bbb01e8fce92bfabb4bd.png"},{"id":96050051,"identity":"c96f56bf-080e-4686-941c-01189af478ec","added_by":"auto","created_at":"2025-11-17 06:36:09","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2866037,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a2e18116ae30ca53731765a4.png"},{"id":96247758,"identity":"23c07069-05b2-4e1b-877c-100376bc772a","added_by":"auto","created_at":"2025-11-19 07:27:42","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15878,"visible":true,"origin":"","legend":"","description":"","filename":"Table5.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c01c47d74e01ebc23b8cb0c2.docx"},{"id":96050065,"identity":"5acb8c4d-6f11-4083-8f1f-8816cc77e5d5","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106844,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/032d8712ea6fbe6c1e10639b.png"},{"id":96050034,"identity":"68a9cfa0-6b42-448e-8320-0e9ca01aea3b","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"docx","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":713847,"visible":true,"origin":"","legend":"","description":"","filename":"Table6.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/781d901c00835082b611d53e.docx"},{"id":96050110,"identity":"9a1593d0-d85d-429a-ae66-995645e32156","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":527572,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/9563ee50abd2e4c196fd1396.png"},{"id":96050028,"identity":"ec74d40b-750a-4d3b-8bf3-8bac3bb000b5","added_by":"auto","created_at":"2025-11-17 06:36:07","extension":"docx","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16510,"visible":true,"origin":"","legend":"","description":"","filename":"Table7.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d532a5d0135ff252268c6ad2.docx"},{"id":96246903,"identity":"422b4f78-ecb9-4f8a-8169-605b1c22e551","added_by":"auto","created_at":"2025-11-19 07:26:51","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":391065,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/90f27c9834dc316b5892e749.png"},{"id":96248153,"identity":"e4a6d365-2c8c-4036-a154-46477b40aedd","added_by":"auto","created_at":"2025-11-19 07:28:06","extension":"docx","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14883,"visible":true,"origin":"","legend":"","description":"","filename":"Table8.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a2b34a2d01a0a9484ad73ed9.docx"},{"id":96050042,"identity":"11b9502b-25f4-46aa-a7e8-b5e661c8ede4","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":388605,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/788c459c543aa41e018a5318.png"},{"id":96050141,"identity":"fe868349-4de8-43a2-87f5-8979a59c895a","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1568310,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/23e9a2c6f53b87e32f218204.png"},{"id":96248208,"identity":"99adc516-b9ea-4feb-ac07-6665ddde7a91","added_by":"auto","created_at":"2025-11-19 07:28:10","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1042429,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/de1937a9028563809c9ad642.png"},{"id":96246038,"identity":"ee570698-495c-46e7-b0b2-533ca7bacf84","added_by":"auto","created_at":"2025-11-19 07:24:18","extension":"docx","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":23664,"visible":true,"origin":"","legend":"","description":"","filename":"Table9.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/39cb2294e3e6338c1d655728.docx"},{"id":96050046,"identity":"3ec4abbf-21fc-4a89-aea5-267d14b337a8","added_by":"auto","created_at":"2025-11-17 06:36:09","extension":"json","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5831,"visible":true,"origin":"","legend":"","description":"","filename":"cc7f085a1434455a948a71230187b40c.json","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ec297502f7c003dbe6298830.json"},{"id":96050055,"identity":"ec2d0693-d44c-4f53-92b0-7e2bd15ae099","added_by":"auto","created_at":"2025-11-17 06:36:09","extension":"docx","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36148,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixA.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4cc451798558c97795e5102b.docx"},{"id":96245686,"identity":"a905054f-06cc-4f9d-85d1-2e3906c3d9ce","added_by":"auto","created_at":"2025-11-19 07:21:51","extension":"xml","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":290211,"visible":true,"origin":"","legend":"","description":"","filename":"cc7f085a1434455a948a71230187b40c1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0d4b68bd91f4cf0ef1213b3f.xml"},{"id":96050097,"identity":"1ce92fdc-7e23-4aa5-aa1f-4bc16be25549","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"jpg","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":529247,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f84da6b59a09d088e6b7c398.jpg"},{"id":96246947,"identity":"7960159a-73e2-46ca-a2b7-104461c09cbe","added_by":"auto","created_at":"2025-11-19 07:26:57","extension":"png","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69907,"visible":true,"origin":"","legend":"","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8f023ea632bb6e89b68b98bc.png"},{"id":96050033,"identity":"9953251b-3055-452a-8e91-d0189085b427","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"png","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4924685,"visible":true,"origin":"","legend":"","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/404b1438287ffeeb1a229eec.png"},{"id":96248118,"identity":"dfb0924b-c49a-4d1a-a307-db79c09ef22f","added_by":"auto","created_at":"2025-11-19 07:28:03","extension":"png","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3547157,"visible":true,"origin":"","legend":"","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0072895a3f59ed00e218e314.png"},{"id":96245829,"identity":"00db251a-7e23-426c-8891-ef1ecc151174","added_by":"auto","created_at":"2025-11-19 07:23:11","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2866037,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7fd3b6e1fadff94839cd4f0a.png"},{"id":96247497,"identity":"576a1546-6f2a-456d-9a31-711176262cb7","added_by":"auto","created_at":"2025-11-19 07:27:31","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106844,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d2ee889f92137dd65290e20b.png"},{"id":96247666,"identity":"e1ebb0bd-7b4d-4f55-9d87-4b343a9cf3c5","added_by":"auto","created_at":"2025-11-19 07:27:40","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":527572,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/861a7a8277a36bda4a499584.png"},{"id":96363391,"identity":"3f5ab1b8-0047-4e8f-9817-e211a6d4ea82","added_by":"auto","created_at":"2025-11-20 10:06:33","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":391065,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/dad4d550898f13c23a2fb617.png"},{"id":96050064,"identity":"ec3efd84-6bb9-4529-a1af-93425f6932f8","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":388605,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7d0a63b5bb550d035b2bcf45.png"},{"id":96245741,"identity":"b7cded27-285b-41d8-b880-9da1cebebcee","added_by":"auto","created_at":"2025-11-19 07:22:17","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1568310,"visible":true,"origin":"","legend":"","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/76a3145484050b656326fec3.png"},{"id":96246259,"identity":"a33f2cd5-fd99-4320-b2a4-394e01bafeaa","added_by":"auto","created_at":"2025-11-19 07:25:13","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1042429,"visible":true,"origin":"","legend":"","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d82ab1ef0d8ed1e8d293f184.png"},{"id":96050073,"identity":"8170efb7-f02f-4d55-8153-0396bb450d52","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"jpeg","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":92948,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6e4a0d2337d3eab2e0cdc022.jpeg"},{"id":96246019,"identity":"a5d6ca9a-a062-4c0b-b3df-8fcb1debaa07","added_by":"auto","created_at":"2025-11-19 07:24:11","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":80286,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/33c18fc0f8078026066b1bd2.png"},{"id":96246499,"identity":"3a845f32-d1d8-493a-86ee-f5531a84fd43","added_by":"auto","created_at":"2025-11-19 07:26:08","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55539,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5929a306f2854a8a1f2a8e43.png"},{"id":96050173,"identity":"c56e127f-13c3-473f-ab6d-c8e76001488f","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"jpeg","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14844,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/188bc9fa1ee3bc29947cb2a0.jpeg"},{"id":96245794,"identity":"98696322-27ad-482d-9065-e343a123abe5","added_by":"auto","created_at":"2025-11-19 07:22:52","extension":"jpeg","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8094,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/419446894e2d5d0cd76ac3e8.jpeg"},{"id":96050069,"identity":"f87f060c-20fd-4883-9bcc-d48994cbe947","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90093,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ad2d64bd8c72d6347bf3908c.png"},{"id":96246882,"identity":"ce3dc145-3edf-45a6-9be8-d00dd1bec415","added_by":"auto","created_at":"2025-11-19 07:26:48","extension":"png","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58218,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/24fe01f629dccc150f248c34.png"},{"id":96050040,"identity":"a6464007-cb6a-49ad-82c7-d60c21ec981c","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"jpeg","order_by":42,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21809,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage16.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/90222f0ed0151274d5b36f63.jpeg"},{"id":96050066,"identity":"b0a05b2e-8098-425a-b6f8-fc529bf07e47","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"jpeg","order_by":43,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20443,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage17.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1c056705f9ed17ed4a1a195d.jpeg"},{"id":96247603,"identity":"32826e8e-3ba9-42b7-9f41-fa4c332a8f78","added_by":"auto","created_at":"2025-11-19 07:27:36","extension":"png","order_by":44,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91980,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage18.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/94a0714b107e8110b7d5df18.png"},{"id":96050136,"identity":"efa327b9-6354-4d62-a2d3-a0771d5c47e6","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":45,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56388,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage19.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e2906453dfd6e648297c838f.png"},{"id":96246116,"identity":"a0e01c53-31f3-4b74-9cb2-dc0a7c3af86c","added_by":"auto","created_at":"2025-11-19 07:24:43","extension":"jpeg","order_by":46,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15935,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b0c6bb0c56cf6446fcc67cf5.jpeg"},{"id":96050048,"identity":"df97ee0b-7577-45be-a00b-1b4ec61f3192","added_by":"auto","created_at":"2025-11-17 06:36:09","extension":"jpeg","order_by":47,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7160,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage20.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a1ffa2065b482abe793131b4.jpeg"},{"id":96050079,"identity":"3007f6e3-f2c9-44de-8457-699d0f141723","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"jpeg","order_by":48,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5318,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage21.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/dac641dab3b1e146d4d6a7a3.jpeg"},{"id":96050118,"identity":"0ebb78bb-688b-4fa1-bdbf-58291870d69d","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"jpeg","order_by":49,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6097,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage22.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7a0248ea39efb9839f8b3e3a.jpeg"},{"id":96050074,"identity":"aed061a9-e8e5-41e9-ac25-d25416d4c185","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":50,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53990,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage23.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/77cf52b5bc5798531ffbc44f.png"},{"id":96050094,"identity":"e44e8878-55ad-485e-8581-dacfe7ce5a1f","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"jpeg","order_by":51,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16544,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage24.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a200851b8f0b688710f3fe84.jpeg"},{"id":96247091,"identity":"78563bf4-09b6-41ad-aa96-d2b73d1fe0bd","added_by":"auto","created_at":"2025-11-19 07:27:08","extension":"jpeg","order_by":52,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7446,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage25.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1d7bb039d4b6d321878c1e2b.jpeg"},{"id":96050045,"identity":"4f961cad-623d-4c68-ad6c-7fe63d932500","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"png","order_by":53,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":606011,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage26.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/38a660c4c9ae516893e388cb.png"},{"id":96050044,"identity":"14bacc16-c82c-41a3-a203-707f100e0a67","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"png","order_by":54,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":387827,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage27.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5b4ab860310503ff9a0d5f97.png"},{"id":96050178,"identity":"38d54aa5-af98-4523-8950-6955dd237fcb","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":55,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106844,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6ddd2497a5a895b2c2835cd8.png"},{"id":96050081,"identity":"9a7f6a77-3264-457e-93cb-19796aaca240","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":56,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":149368,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage29.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/20b4f5fc3369418c6aa770d2.png"},{"id":96050138,"identity":"d0175215-6a99-43f5-bcfc-16be5cf1a464","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":57,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53303,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/53610ffcfb0c857103efa844.png"},{"id":96050120,"identity":"b8e23642-6b66-4a5e-a6f5-1ba3317f9c08","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":58,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":391065,"visible":true,"origin":"","legend":"","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/db744054928ab5fe551f5b7d.png"},{"id":96247525,"identity":"33d91cf2-3f76-479b-8550-104de94054e3","added_by":"auto","created_at":"2025-11-19 07:27:32","extension":"png","order_by":59,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":388605,"visible":true,"origin":"","legend":"","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/efb6a597c198ca3b69bad463.png"},{"id":96050086,"identity":"aa9eabf0-e919-4d9b-a51a-5d27baeb106c","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"png","order_by":60,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":177376,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage32.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e4b00b9e5649170328718b29.png"},{"id":96050160,"identity":"ab741cce-ad55-49d0-846c-672ac0971f1b","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"jpeg","order_by":61,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5840,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage33.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/34d42304b37c5f053c891438.jpeg"},{"id":96050061,"identity":"95e9a3c1-d9ff-4d79-b144-dada416ab41d","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":62,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62638,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage34.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7951db0be82b42c51065779a.png"},{"id":96050077,"identity":"403b0fff-ed7c-4cdc-b447-c252efba11d8","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":63,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53832,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage35.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/835dd0db0acea144416cd81f.png"},{"id":96248029,"identity":"5d671125-a7c2-430c-b118-d7a4ca3ad64b","added_by":"auto","created_at":"2025-11-19 07:27:58","extension":"png","order_by":64,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82059,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/995039f69e0d9fa04d817909.png"},{"id":96050088,"identity":"8a333803-3c0a-42ac-8a55-92e555926db5","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"jpeg","order_by":65,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9837,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage37.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0a224f2d0ed73723c639ce7a.jpeg"},{"id":96247702,"identity":"2ae8bf92-8b45-4efa-bc5a-331162a210ab","added_by":"auto","created_at":"2025-11-19 07:27:41","extension":"png","order_by":66,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54155,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage38.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cbcfdfc597a31a5a322731c0.png"},{"id":96050134,"identity":"58a77c86-11bc-446e-a794-5d233f484bd4","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":67,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49696,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage39.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/51b733e917a7010f70e144c6.png"},{"id":96246104,"identity":"33c5617a-df42-4b59-ab5b-561e47ee4b4b","added_by":"auto","created_at":"2025-11-19 07:24:39","extension":"jpeg","order_by":68,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21415,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/90d7381d9a4525f52264c04f.jpeg"},{"id":96246030,"identity":"5fc07f62-54d6-4e7c-8f6e-9169e1456e04","added_by":"auto","created_at":"2025-11-19 07:24:17","extension":"png","order_by":69,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63186,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage40.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4ba52c76e16465b56cdf9be0.png"},{"id":96245726,"identity":"0a0c02a7-3b51-416d-be4b-b5bbbb8d852d","added_by":"auto","created_at":"2025-11-19 07:22:08","extension":"jpeg","order_by":70,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6262,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage41.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/aef677ab917bd1b67e4aea58.jpeg"},{"id":96248145,"identity":"0cf862b6-290a-486c-bf23-83f7871af632","added_by":"auto","created_at":"2025-11-19 07:28:06","extension":"png","order_by":71,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90753,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage42.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d628df429f750767e79c088c.png"},{"id":96248068,"identity":"118a9337-18cf-4b23-a921-a8b5ec577fa0","added_by":"auto","created_at":"2025-11-19 07:28:00","extension":"png","order_by":72,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55800,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage43.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/334f9f2b44cee455120c92d2.png"},{"id":96246536,"identity":"b245c4e5-91f6-4221-90d9-042884b2ff98","added_by":"auto","created_at":"2025-11-19 07:26:11","extension":"png","order_by":73,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30323,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage44.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/462a6a8ca0e2c964bcf00301.png"},{"id":96050109,"identity":"4254a635-e5e5-44d4-a485-23336f7bc883","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":74,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64607,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage45.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/2b77136e5651f5d6cbfc8295.png"},{"id":96050053,"identity":"e39cb340-20e9-4335-ba3e-ba76c75e1daf","added_by":"auto","created_at":"2025-11-17 06:36:09","extension":"png","order_by":75,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":67867,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage46.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/20ee8d8e1c53f9f217cc273b.png"},{"id":96050092,"identity":"eb9bceee-ba22-43b0-b089-f992c65ced73","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"png","order_by":76,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":159353,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage47.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d22e0b1ec0833c66d54f8275.png"},{"id":96050177,"identity":"f237a2d6-9ab0-48aa-bee7-a2122fb4b470","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":77,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":69907,"visible":true,"origin":"","legend":"","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5bcd36b0d50b03fe95021b23.png"},{"id":96050189,"identity":"65246fa9-1aa3-4feb-ada2-be351058beeb","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":78,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":747375,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage49.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/43b073e31e75b98c5f81e352.png"},{"id":96050084,"identity":"4cb4ee78-e14f-4ac3-9d5d-781288004556","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"jpeg","order_by":79,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19363,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1611e016698e4e95e04da2d7.jpeg"},{"id":96245697,"identity":"209f8360-9684-4151-96d2-56248764c50a","added_by":"auto","created_at":"2025-11-19 07:21:58","extension":"jpeg","order_by":80,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15935,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8ec1664d85b3ab94dabace58.jpeg"},{"id":96050089,"identity":"46eea717-d883-47a9-ab80-2ec8088db0b5","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"png","order_by":81,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53303,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f7b2ebecbf56f634cffd3f31.png"},{"id":96050039,"identity":"cdb48020-8702-489d-ba04-d80fdc473664","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"jpeg","order_by":82,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21415,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a874f5d005e60583cf483c1f.jpeg"},{"id":96050121,"identity":"c1aaec6a-86b4-47f3-b977-c941033b414c","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"jpeg","order_by":83,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19363,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b5654a7a629791821faac848.jpeg"},{"id":96050159,"identity":"e4b7fb64-7eea-467f-8af8-b85f021ff2a9","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":84,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82227,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage54.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/fed1be5cf54fef3a17b99f19.png"},{"id":96050127,"identity":"13811f53-c55d-4058-89ea-607fd218efad","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":85,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82059,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/23b9824f5beaa04161c0f3f9.png"},{"id":96050142,"identity":"255022d7-bda6-4e3c-8bde-17855c290752","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"jpeg","order_by":86,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6611,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage56.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5ad22faaaefb57b6ae897341.jpeg"},{"id":96247970,"identity":"482db046-7aca-40e3-b231-5191e44bd4e3","added_by":"auto","created_at":"2025-11-19 07:27:56","extension":"jpeg","order_by":87,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6605,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage57.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0f128ce3dee3adef60dff4ee.jpeg"},{"id":96050100,"identity":"87b787da-76cf-4b0b-a3cc-23dce5979b28","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":88,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":80286,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/659e55c8c896576573463ad7.png"},{"id":96050154,"identity":"59ab5b31-4c8b-441b-954c-5f6f4a8d5820","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":89,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55539,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/43fde747fdfb9d442e3cedea.png"},{"id":96050202,"identity":"99515e0e-a15b-48c4-9ee4-b6c0103e7e5c","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":90,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82227,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage54.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/656577b95ee6c1aa1aef7ff3.png"},{"id":96247492,"identity":"48afbaf6-9b73-4328-b311-42def8aae481","added_by":"auto","created_at":"2025-11-19 07:27:31","extension":"jpeg","order_by":91,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14844,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage12.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/62b45173900a51b2bf2ea64c.jpeg"},{"id":96246678,"identity":"7d3e15b1-4ecd-4268-a66f-6ef573f7fc6a","added_by":"auto","created_at":"2025-11-19 07:26:30","extension":"jpeg","order_by":92,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8094,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage13.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6fc7d0b66378c2258f452e27.jpeg"},{"id":96246128,"identity":"a9d672f0-d74d-493e-9044-6764d062b7f0","added_by":"auto","created_at":"2025-11-19 07:24:44","extension":"png","order_by":93,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90093,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/34ff4856ff6a72b6103e88ea.png"},{"id":96246431,"identity":"d5a72aff-729a-47bc-ac2a-51252e21d5f3","added_by":"auto","created_at":"2025-11-19 07:25:59","extension":"png","order_by":94,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58218,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8d43847ffe9624555a21d699.png"},{"id":96050183,"identity":"c2e56399-1df1-4d3e-b746-9f21b3c53903","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"jpeg","order_by":95,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":21809,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage16.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6872f9fa671f046c5675a371.jpeg"},{"id":96247618,"identity":"33591190-f9be-4fab-ab9e-c55b51c965d5","added_by":"auto","created_at":"2025-11-19 07:27:37","extension":"jpeg","order_by":96,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20443,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage17.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5809f8def25fc43abdc91055.jpeg"},{"id":96247528,"identity":"f96d4ce5-e500-4142-b8ff-61db395b24db","added_by":"auto","created_at":"2025-11-19 07:27:32","extension":"png","order_by":97,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":91980,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage18.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6ebf89592736756720e3efa8.png"},{"id":96245877,"identity":"c229a9b0-6686-416f-8cfc-6bcd5b89427a","added_by":"auto","created_at":"2025-11-19 07:23:24","extension":"png","order_by":98,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":56388,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage19.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/944e0e191ed2e2c3386c7fdb.png"},{"id":96050171,"identity":"410b93f1-5916-4410-9d2d-a492562cf447","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"jpeg","order_by":99,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7160,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage20.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/23eccbb33a71cb085d7b3593.jpeg"},{"id":96050104,"identity":"68b91c46-b411-4120-9511-4d56cf7b3fdb","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"jpeg","order_by":100,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5318,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage21.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/14fb5d7400f422106724c7bc.jpeg"},{"id":96050056,"identity":"66916007-59a4-4732-b6e9-d8ac65c299d0","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":101,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82059,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/fb0d8f1adc3138bc749cd3a0.png"},{"id":96247566,"identity":"4c0537b1-a374-4688-bf35-773d963aff74","added_by":"auto","created_at":"2025-11-19 07:27:34","extension":"jpeg","order_by":102,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6097,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage22.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b9466f789a63117cc083ed73.jpeg"},{"id":96247163,"identity":"ed8526df-d765-4683-9727-c6f6ebc2feec","added_by":"auto","created_at":"2025-11-19 07:27:13","extension":"png","order_by":103,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53990,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage23.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b804aaf4f1b5cc5bdcb82c94.png"},{"id":96050116,"identity":"37cabd93-9110-4c79-86ca-bc704c96abb6","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"jpeg","order_by":104,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16544,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage24.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ca0e7a4b83289b720e046e7b.jpeg"},{"id":96050167,"identity":"fafa4ec6-c5e9-4475-926e-7ff5e33a08c4","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"jpeg","order_by":105,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7446,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage25.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4a53056afd4348bab2754f12.jpeg"},{"id":96247539,"identity":"0190196b-05a4-4d1d-9e21-81047cf2dd61","added_by":"auto","created_at":"2025-11-19 07:27:33","extension":"jpeg","order_by":106,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5840,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage33.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5692f0185b2bb1e2efce16f8.jpeg"},{"id":96050067,"identity":"88d25aba-fd24-42d3-ba52-93bc0667ac8d","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":107,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62638,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage34.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b671cd238dc13eaf28111a14.png"},{"id":96247431,"identity":"dd6b431e-41f8-490a-8907-16cc22e1a672","added_by":"auto","created_at":"2025-11-19 07:27:28","extension":"png","order_by":108,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":53832,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage35.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d8c299bde3b1acc628d0fc67.png"},{"id":96050248,"identity":"8c1eb675-2092-4a85-9476-e68f6c16c57a","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":109,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":82059,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/48e62684a1eff09c962a0a86.png"},{"id":96247947,"identity":"a2763fd4-2627-4b9b-af25-b0d2d54a865a","added_by":"auto","created_at":"2025-11-19 07:27:54","extension":"jpeg","order_by":110,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9837,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage37.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/03333ab706733323fa44fdba.jpeg"},{"id":96050238,"identity":"92a3e283-e39b-4e97-9131-9ab2616749af","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":111,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54155,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage38.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/40bd04d92d5e738bf8499a96.png"},{"id":96247950,"identity":"68ba496c-c1b8-4ed6-9d28-d6ca0cb998e1","added_by":"auto","created_at":"2025-11-19 07:27:54","extension":"jpeg","order_by":112,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6611,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage56.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7b5a6640f3e0a3f8c8fef07a.jpeg"},{"id":96050165,"identity":"502afc06-b806-469f-afee-fe7f244eaec6","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":113,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49696,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage39.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/9d0560ab41b00eb948da7efd.png"},{"id":96247005,"identity":"0f01ba8a-3658-442a-928e-9c240da24811","added_by":"auto","created_at":"2025-11-19 07:27:02","extension":"png","order_by":114,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63186,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage40.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8e76d62ec903b383c9689e46.png"},{"id":96246226,"identity":"aafc4a26-8edf-4591-9c75-f052108b44d3","added_by":"auto","created_at":"2025-11-19 07:25:06","extension":"jpeg","order_by":115,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6262,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage41.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c02e296123cde8e047c13fd6.jpeg"},{"id":96248346,"identity":"d5fbb4f6-b312-4ba8-815b-7953dfaf7eac","added_by":"auto","created_at":"2025-11-19 07:28:21","extension":"png","order_by":116,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90753,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage42.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ce8fa0cda12327cc85229258.png"},{"id":96247303,"identity":"8b7e307a-7b72-4bb9-8547-3058ff1aee6c","added_by":"auto","created_at":"2025-11-19 07:27:21","extension":"png","order_by":117,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":55800,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage43.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/30422a28e57b09995321d7f4.png"},{"id":96050181,"identity":"cc2dddf2-ca2b-4693-8809-1100c63a7116","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":118,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":30323,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage44.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/53967353f7861dc967c4cd70.png"},{"id":96050231,"identity":"3b92b1e1-2ec9-4b81-8818-6a13168a8e1a","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":119,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64607,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage45.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e7cb94dedc6518fe5ee1ab28.png"},{"id":96246146,"identity":"398482f3-8430-428d-8f68-cb374127a5e7","added_by":"auto","created_at":"2025-11-19 07:24:45","extension":"png","order_by":120,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":67867,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage46.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/110064b770084f42b29ddee9.png"},{"id":96050200,"identity":"36be00d2-08bc-42ce-ab4b-63f0edcf47e1","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"jpeg","order_by":121,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6605,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage57.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/956e2f40461e310655273775.jpeg"},{"id":96248072,"identity":"09063d33-6269-4865-9913-bc71a49713c4","added_by":"auto","created_at":"2025-11-19 07:28:01","extension":"wmf","order_by":122,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1668,"visible":true,"origin":"","legend":"","description":"","filename":"image39.wmf","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a668a39e0e87217bba60d9d9.wmf"},{"id":96050203,"identity":"fe235557-b383-423d-960c-d1397d5ef862","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":123,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64675,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/3d96cd263a32089f22443760.png"},{"id":96050085,"identity":"7ec71113-c143-4091-ae48-27b5c83682ec","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"png","order_by":124,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19425,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b39ea18e3419ae7bede520bf.png"},{"id":96050230,"identity":"aad33aaa-cd66-45c5-8d11-25a441dbe6a9","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":125,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1003469,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/bdb5d777888d9d530f4bc96d.png"},{"id":96050262,"identity":"1b80bb2b-63ec-4f29-94e4-5bae2a3fdcd8","added_by":"auto","created_at":"2025-11-17 06:36:20","extension":"png","order_by":126,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":743025,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/2d407cdd385a250b52cc0db8.png"},{"id":96050236,"identity":"472deb76-5ff0-4a87-bdc3-c2aae6b5761f","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":127,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":618160,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/65d922f48fa63e9aa568cb50.png"},{"id":96050130,"identity":"8df152c2-f28b-4a6b-844b-d7df2ea402b1","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":128,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":32522,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c36049c357c03a41fae250bc.png"},{"id":96248470,"identity":"980849b5-74fb-49ba-b7a9-5fcf95f81dff","added_by":"auto","created_at":"2025-11-19 07:28:29","extension":"png","order_by":129,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":146824,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1b3263e6e57d9e65961a0e42.png"},{"id":96050103,"identity":"26b1bcca-763a-4175-998b-f9a3df948d9b","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":130,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90256,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ef8a40a7fa2e318253bf8687.png"},{"id":96050124,"identity":"078edd19-5038-4a05-b2e9-e4517a78d9a0","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":131,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89666,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/948fe1b6a1d15e824bae1b52.png"},{"id":96050246,"identity":"f1745fda-36e2-4cf1-98aa-df893af3a8fa","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":132,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":602405,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/b37cdfa38406412dc07c3675.png"},{"id":96050062,"identity":"04a051b3-8b0e-46d2-bba2-c2df79def160","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":133,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":353482,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cf0f5d214bd8f676ff58846a.png"},{"id":96248359,"identity":"f3699c06-7610-48b5-83a6-7d250780b34d","added_by":"auto","created_at":"2025-11-19 07:28:22","extension":"png","order_by":134,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":85366,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c3745a5f51774719b38bfa4f.png"},{"id":96050228,"identity":"5d327c2e-6d32-4b08-9d11-435204863dae","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":135,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":23878,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/ceb1adb7572afdce1d7c8923.png"},{"id":96246138,"identity":"8b538e3a-0a1a-43ab-8ff1-18ae9d6ca538","added_by":"auto","created_at":"2025-11-19 07:24:44","extension":"png","order_by":136,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12919,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8cebe95a76f4c7937c01ede9.png"},{"id":96050194,"identity":"d892ddab-1c3a-4634-86d4-61e2103369f6","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":137,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15043,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage12.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/92bceb0761623fb340d1d0c0.png"},{"id":96050161,"identity":"ba236747-e99b-4528-a830-dc9dae031871","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":138,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13579,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage13.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/dab15c574effdd4b2bba3252.png"},{"id":96050250,"identity":"1de26ff6-23f9-415e-ba74-37c15b7d6f78","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":139,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19070,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage14.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d7bce76db503669c9c0dd5f3.png"},{"id":96246072,"identity":"37b97744-212d-4e4c-8c3f-7d8a9a79890b","added_by":"auto","created_at":"2025-11-19 07:24:30","extension":"png","order_by":140,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13333,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage15.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5aae83d49ff58d74895d5886.png"},{"id":96246958,"identity":"fad6edf2-d410-44d2-bce5-bea81ea6689a","added_by":"auto","created_at":"2025-11-19 07:26:57","extension":"png","order_by":141,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16278,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage16.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/dec99dce934c26636bea1d28.png"},{"id":96247211,"identity":"b6869e12-fa89-4d75-b4cc-6a57583d1abe","added_by":"auto","created_at":"2025-11-19 07:27:15","extension":"png","order_by":142,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14934,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage17.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/01ba9f103b8ce9f0003e333c.png"},{"id":96247080,"identity":"e1ed55f5-cb39-45a4-a15a-5f7e0b041a3d","added_by":"auto","created_at":"2025-11-19 07:27:07","extension":"png","order_by":143,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12579,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage18.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/71f6db4bc036d20ca3dbe0ef.png"},{"id":96050137,"identity":"54697fd8-01e4-4332-9819-c903119841bc","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":144,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13273,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage19.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4e4f9c733f9dee188912eb24.png"},{"id":96247939,"identity":"02b93bcd-6c42-4a62-9a0a-e0102de884af","added_by":"auto","created_at":"2025-11-19 07:27:52","extension":"png","order_by":145,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17378,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4e2cf219f3d8871c9c7c7d2b.png"},{"id":96050148,"identity":"41b0e2eb-daa7-4b8b-865b-cd9d25cca770","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":146,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12204,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage20.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/51b07e2be0284b2d39cc05a2.png"},{"id":96050190,"identity":"a266628c-d927-4784-ad78-a2b10e64b889","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":147,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12768,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage21.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6d41131798d3e852d7bc5992.png"},{"id":96050123,"identity":"82a96573-a961-47cb-b20d-ca21fff32710","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":148,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17950,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage22.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8b74ae850441553f008fad0c.png"},{"id":96247052,"identity":"7635d4ee-26c4-46d4-a105-0b655c8854fe","added_by":"auto","created_at":"2025-11-19 07:27:05","extension":"png","order_by":149,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12232,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage23.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/bbf835c61fcd18dcad52ddba.png"},{"id":96050245,"identity":"c836fee8-db7c-40ea-b040-c337b82e0139","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":150,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11843,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage24.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/9e3234df37db61ec70462304.png"},{"id":96050268,"identity":"0147d025-e2ca-4ab7-b0a9-4181463ab8cd","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":151,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11465,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage25.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/652e11baae350d0f607e567d.png"},{"id":96246271,"identity":"d7021b27-7b69-48b1-a7b7-050e6dd9f1ed","added_by":"auto","created_at":"2025-11-19 07:25:14","extension":"png","order_by":152,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":132992,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage26.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/32844f669f7e134facaf7938.png"},{"id":96050265,"identity":"9d7941c7-56b8-4f6e-9f3a-4f3937e89a6c","added_by":"auto","created_at":"2025-11-17 06:36:20","extension":"png","order_by":153,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83086,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage27.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6a5289327f371abedddb2f04.png"},{"id":96050249,"identity":"eb12f1d0-c987-4229-95e5-6d79508e1f87","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":154,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":32522,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage28.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1654662fa5520e7962bd9df4.png"},{"id":96050196,"identity":"6717c16d-1485-4aae-9121-f26900685025","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":155,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":37705,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage29.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/48302930201dc50c04d56f11.png"},{"id":96248143,"identity":"b458805a-52c1-454a-8fd6-0f869075c27d","added_by":"auto","created_at":"2025-11-19 07:28:06","extension":"png","order_by":156,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11684,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/be1d70f1d7f0082f8e4ba6ca.png"},{"id":96050133,"identity":"6e395d18-3f60-4953-b256-c756e149650e","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":157,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":90256,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage30.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/04fb97af564944b2011e89e5.png"},{"id":96246235,"identity":"4b6ab047-48e8-4d73-a118-ca5ffbd33dcb","added_by":"auto","created_at":"2025-11-19 07:25:08","extension":"png","order_by":158,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89666,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage31.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/862c9b168ece983a9aa0ac1f.png"},{"id":96050083,"identity":"8f593e51-b2a1-48f6-a34f-1768e3a3adfa","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":159,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49797,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage32.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8f3d5dc2ec5f6dbaabf55967.png"},{"id":96050143,"identity":"3b0aa345-07f1-4675-a3c6-d90c40dabd4e","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":160,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8536,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage33.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c02ed2f801cc2d76caf2fabf.png"},{"id":96050145,"identity":"43bbda40-d313-4eac-ba64-348f427801fa","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":161,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12192,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage34.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/558cd93400b0f918add70812.png"},{"id":96050082,"identity":"cca7c3a3-5d5a-469d-85d5-e7160f0da2c6","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":162,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8493,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage35.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/c920ad7bd78b6ec17468f4bf.png"},{"id":96050195,"identity":"b134e5d9-25f5-4ab4-96d2-c62b163641b6","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":163,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17147,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/3934bcae2e0ff302ded03cca.png"},{"id":96050242,"identity":"d86c762d-f71e-4653-9eb6-a03737a6119a","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":164,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15604,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage37.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/593fcd26e483fd6918a427e2.png"},{"id":96246123,"identity":"ba8c636f-7474-4263-8105-8a6d5dcc48ce","added_by":"auto","created_at":"2025-11-19 07:24:43","extension":"png","order_by":165,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14831,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage38.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e7ee83d0f21f9c1e7d4d8ea0.png"},{"id":96245977,"identity":"5eb22161-9f9e-47c7-b458-61085f3b7463","added_by":"auto","created_at":"2025-11-19 07:24:02","extension":"png","order_by":166,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8275,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage39.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/4a0fcd34ac33a2378af69bf0.png"},{"id":96050270,"identity":"c23c3fdd-ca86-4b47-8eae-24a1fcfffd06","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":167,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15598,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/283f4f54b6f7241762e0677a.png"},{"id":96246865,"identity":"ad9f6697-4038-4fc6-9b0f-6d2abd3847ad","added_by":"auto","created_at":"2025-11-19 07:26:47","extension":"png","order_by":168,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13344,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage40.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cb5ec4914ccd368e171ce393.png"},{"id":96248103,"identity":"b4706c90-04a7-415f-9c7d-584c9f8c70c8","added_by":"auto","created_at":"2025-11-19 07:28:03","extension":"png","order_by":169,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9963,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage41.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8b9914b146b846d2d426eb94.png"},{"id":96050131,"identity":"9e0e60af-3bc0-4af8-afac-c24c7cab2252","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":170,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11139,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage42.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/22caaa03e8ffeb90a7f699af.png"},{"id":96050234,"identity":"35702e69-54f1-40bb-bdcb-d600a26eb997","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":171,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9954,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage43.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/775fadaeee2ec256d56d1a76.png"},{"id":96246069,"identity":"8061a68d-9909-45f9-8dff-1d86c0f615d8","added_by":"auto","created_at":"2025-11-19 07:24:29","extension":"png","order_by":172,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9293,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage44.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/af6c78bb0d47843c7683e90c.png"},{"id":96050106,"identity":"c27f68b8-e250-4de1-8490-1c17e56bf2c3","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":173,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11632,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage45.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/587e224fd4cf928562587380.png"},{"id":96246009,"identity":"f15df25e-3ac2-43a7-986c-d55b04edacb1","added_by":"auto","created_at":"2025-11-19 07:24:08","extension":"png","order_by":174,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11664,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage46.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/fb5c05d42c9d555321f36a2e.png"},{"id":96248052,"identity":"722c81b5-b951-4b8c-ac2d-c54e2f2fe35c","added_by":"auto","created_at":"2025-11-19 07:27:59","extension":"png","order_by":175,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":52858,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage47.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6e6d61c130b502069b03186f.png"},{"id":96050135,"identity":"a1619be0-e84e-4208-a924-45c5d1ff35e0","added_by":"auto","created_at":"2025-11-17 06:36:14","extension":"png","order_by":176,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19425,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage48.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/19a42026d746ddaca81c071c.png"},{"id":96050075,"identity":"2dcbcad4-c46c-4f0c-bd2e-5ee92e6b0699","added_by":"auto","created_at":"2025-11-17 06:36:11","extension":"png","order_by":177,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":163987,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage49.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/83ecc1680c1c69bc22d9aba5.png"},{"id":96050261,"identity":"fdc14c17-87c5-4389-ab47-f12fc50c66f5","added_by":"auto","created_at":"2025-11-17 06:36:20","extension":"png","order_by":178,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14691,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5279f5ce82f24bd1f1e89a6f.png"},{"id":96247761,"identity":"d6c95e0d-cc0c-466d-9a10-0c3de7bbb0bd","added_by":"auto","created_at":"2025-11-19 07:27:42","extension":"png","order_by":179,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17378,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage50.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5378af0d628a564c45a20734.png"},{"id":96246334,"identity":"6fe19ead-542b-4477-845d-19c7f1b4d58b","added_by":"auto","created_at":"2025-11-19 07:25:27","extension":"png","order_by":180,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11684,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/68d523dd3289f2185787c5c6.png"},{"id":96050258,"identity":"198110c1-9e2d-4120-9948-dd5bba190bc1","added_by":"auto","created_at":"2025-11-17 06:36:20","extension":"png","order_by":181,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15598,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage52.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/331de98e033694f8e6aededb.png"},{"id":96050255,"identity":"bdd33408-6087-463a-8515-8014e2e49829","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":182,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14691,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage53.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f4ba257513e94959f5f97679.png"},{"id":96245781,"identity":"89d4bc4a-1c67-45a5-9d56-363d589351fd","added_by":"auto","created_at":"2025-11-19 07:22:36","extension":"png","order_by":183,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17239,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage54.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e7c737193ca90c65fd0da9c0.png"},{"id":96246174,"identity":"0d171c74-7b08-4449-9107-5169893a3dbb","added_by":"auto","created_at":"2025-11-19 07:24:56","extension":"png","order_by":184,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17147,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage36.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/566ffcd6acfa57b90eb9cb49.png"},{"id":96050263,"identity":"256f87a2-128f-434f-a1e1-105b37cca8f5","added_by":"auto","created_at":"2025-11-17 06:36:20","extension":"png","order_by":185,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11219,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage56.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/78bcef3834ece82745fe436f.png"},{"id":96050163,"identity":"4ff71d33-0305-4e4f-83e2-af85e0b7be58","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":186,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9989,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage57.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/40b2c00ecd5da93c1863ff6c.png"},{"id":96247897,"identity":"55334950-9ff7-426a-ad3a-bfbbe7974ede","added_by":"auto","created_at":"2025-11-19 07:27:49","extension":"png","order_by":187,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":23878,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage58.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/73ce215485a11ba0a5cd4c7a.png"},{"id":96247460,"identity":"536b8433-38d8-4c35-9416-5e69ca8763fa","added_by":"auto","created_at":"2025-11-19 07:27:29","extension":"png","order_by":188,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12919,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage59.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f5ef70578a63353dc519af64.png"},{"id":96050185,"identity":"e565284e-4b40-4fe6-8d8a-2f5e3baef6dc","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":189,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17239,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a6afaeb09676c9a54bbdc3ed.png"},{"id":96050192,"identity":"6718875c-122d-4aad-844d-ed240cf55ac6","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":190,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15043,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage60.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/206b18ae06a4a8e06aa68ce6.png"},{"id":96245716,"identity":"8d66acfe-ce1f-40af-a7f1-5729ece10815","added_by":"auto","created_at":"2025-11-19 07:22:04","extension":"png","order_by":191,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13579,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage61.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/f625d864f30c5bc6ff5ca52c.png"},{"id":96247504,"identity":"8effb0d9-5038-4a07-9a5a-b47ae3d5568f","added_by":"auto","created_at":"2025-11-19 07:27:32","extension":"png","order_by":192,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19070,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage62.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a27aef6279c33a9834bc78ff.png"},{"id":96050244,"identity":"fdb795ab-b242-4bcf-8147-47cc7263402c","added_by":"auto","created_at":"2025-11-17 06:36:19","extension":"png","order_by":193,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13333,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage63.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/9de135869ef3fd12423d698f.png"},{"id":96050180,"identity":"6ceee987-6e19-41db-9bf6-a8a0cfd9da85","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":194,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16278,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage64.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1c5bc5e0bc7e8873992156d2.png"},{"id":96050166,"identity":"a5f4467b-db74-44d2-b06b-5c0878fa1eab","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":195,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14934,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage65.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/62f132a75d64c3ce6bcd1be0.png"},{"id":96246206,"identity":"99a4107e-7f67-493b-b729-6fd545e8e0f8","added_by":"auto","created_at":"2025-11-19 07:25:05","extension":"png","order_by":196,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12579,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage66.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/28dcdc9a143ffc59e37e90cd.png"},{"id":96050152,"identity":"39b8c735-435a-4213-a52d-6e0e17d5a9a3","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":197,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13273,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage67.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/3a4d25cf17e2a700d331b2da.png"},{"id":96050226,"identity":"5f0fd5a7-0aaf-4ff0-87fd-d55a9a970449","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":198,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12204,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage68.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/13e3b5d4f4773cfcc4fd80fc.png"},{"id":96050157,"identity":"1c906e00-9579-4054-9e23-5506e07cc3ab","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":199,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12768,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage69.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/3c258849e6344730c4e45908.png"},{"id":96245798,"identity":"0f5c74eb-3796-4e70-a79c-656484786348","added_by":"auto","created_at":"2025-11-19 07:22:55","extension":"png","order_by":200,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17147,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/60d4e6b8bdb2b271239ca973.png"},{"id":96050271,"identity":"7ef574cb-4bcf-4723-8f5c-8f301d2c5fee","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":201,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17950,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage70.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/89e9e245908fd86fd005199e.png"},{"id":96050156,"identity":"43d1d1d6-0587-43e2-b797-32a1e761b7c0","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":202,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12232,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage71.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/fa6f4f65bcae418be5f8cb61.png"},{"id":96050111,"identity":"686d8eeb-f661-4dbd-81bc-4464f6e6622b","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":203,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11843,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage72.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/02c219cd303ba2ab2d2090ac.png"},{"id":96248467,"identity":"dcf6be3b-f2df-4d72-9331-8b09ad280ebc","added_by":"auto","created_at":"2025-11-19 07:28:29","extension":"png","order_by":204,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11465,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage73.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/af27e8ad7e30b99c6367caba.png"},{"id":96050213,"identity":"d1bbcb8d-42ee-4a07-91b8-eed2456c448a","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":205,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8536,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage33.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d91e7deec297b5698a534482.png"},{"id":96246300,"identity":"c468632e-5409-4988-a83c-73756d3bda6a","added_by":"auto","created_at":"2025-11-19 07:25:21","extension":"png","order_by":206,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12192,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage34.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/e56596bebe28dea8a14e81d5.png"},{"id":96050267,"identity":"a4467d14-c570-4cf6-8a0f-20c3dd85cbec","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":207,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8493,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage76.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/cb5b2ea7334827c886f27533.png"},{"id":96050199,"identity":"87c3f2e6-d391-4622-befe-c4fdac3239c8","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":208,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17147,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/d304275b0a091fb7dd8c089a.png"},{"id":96247189,"identity":"d10611ae-8a00-4467-b4e0-d39603b15ca0","added_by":"auto","created_at":"2025-11-19 07:27:14","extension":"png","order_by":209,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15604,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage78.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/29f078b146fc56ea518a9e7d.png"},{"id":96247072,"identity":"178205f8-7581-4aec-b057-0a16e43c14ab","added_by":"auto","created_at":"2025-11-19 07:27:07","extension":"png","order_by":210,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14831,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage79.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/91225ee37f6285db6a342cee.png"},{"id":96247637,"identity":"3dabdd12-00ae-4be0-800a-93225a9b136e","added_by":"auto","created_at":"2025-11-19 07:27:38","extension":"png","order_by":211,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11219,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/5d862b0a25271b22ae7e890a.png"},{"id":96050193,"identity":"3aa6daa1-5e7d-429e-9556-f7d63a4e8161","added_by":"auto","created_at":"2025-11-17 06:36:17","extension":"png","order_by":212,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8275,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage80.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7e0945d1de0b4044742b837c.png"},{"id":96247035,"identity":"28983143-1d26-4f6e-b475-5ed2a862269a","added_by":"auto","created_at":"2025-11-19 07:27:04","extension":"png","order_by":213,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13344,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage81.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/69b2a235b5258bf5814ec055.png"},{"id":96050115,"identity":"1cf7b9df-0a4a-48af-aa8f-212b05b25580","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":214,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9963,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage82.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/08de61a52b27d8183e64c48a.png"},{"id":96050187,"identity":"2e63667b-0534-43dd-85f3-5b7431efa675","added_by":"auto","created_at":"2025-11-17 06:36:16","extension":"png","order_by":215,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11139,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage83.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6a518372625e3846c8ae32a2.png"},{"id":96050239,"identity":"695054fa-118e-452d-92a6-428983f88c6e","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":216,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9954,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage84.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/47b675a062162ad37a8614e7.png"},{"id":96050235,"identity":"734d2c74-d07f-4799-98c0-b05de4749e22","added_by":"auto","created_at":"2025-11-17 06:36:18","extension":"png","order_by":217,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9293,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage44.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/98012ed194c67b43d46a1c61.png"},{"id":96050269,"identity":"4492acd3-b588-4458-bde5-a61ad3472881","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":218,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11632,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage86.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/08ccaa89f9711c244cf86a8e.png"},{"id":96050274,"identity":"9583f669-8619-497d-aca9-9ac008824c2d","added_by":"auto","created_at":"2025-11-17 06:36:21","extension":"png","order_by":219,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11664,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage87.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8d9b63cb98809ce3b3abf29f.png"},{"id":96050150,"identity":"1ae1036e-652f-4f15-84f9-b7f8b4ad0542","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"png","order_by":220,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9989,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage57.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/868254aaacc0e8cc8904b7cb.png"},{"id":96245711,"identity":"9af6b07e-5164-4d5d-94c9-81ccf29d418a","added_by":"auto","created_at":"2025-11-19 07:22:01","extension":"png","order_by":221,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1806,"visible":true,"origin":"","legend":"","description":"","filename":"Onlineimage39.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/bcb9ed71179d36a58d319ae0.png"},{"id":96050164,"identity":"f3f048df-c0d3-4c33-a7f6-048503c54717","added_by":"auto","created_at":"2025-11-17 06:36:15","extension":"xml","order_by":222,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":289227,"visible":true,"origin":"","legend":"","description":"","filename":"cc7f085a1434455a948a71230187b40c1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/1edcc9a6f3690d5476ffcc75.xml"},{"id":96050275,"identity":"9479e357-b2aa-4857-b304-d27b359c986b","added_by":"auto","created_at":"2025-11-17 06:36:22","extension":"html","order_by":223,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":298839,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/86e7a78b0dd101b3c3f407eb.html"},{"id":96247613,"identity":"672f4190-0809-4f7f-84ba-f45ad29bf93e","added_by":"auto","created_at":"2025-11-19 07:27:37","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":529247,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution map of CTSs.\u003c/p\u003e","description":"","filename":"Figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/66aa98c9ce4948dafec61bb0.jpg"},{"id":96247865,"identity":"07d3ddd0-75ad-4dbf-aaf8-71c93065e3d4","added_by":"auto","created_at":"2025-11-19 07:27:48","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":3547157,"visible":true,"origin":"","legend":"\u003cp\u003eLandscape elements of traditional architecture.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/2e1fbad90b71bdce854bf3d0.png"},{"id":96245669,"identity":"e2005251-3810-49ad-876f-193828601217","added_by":"auto","created_at":"2025-11-19 07:21:44","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":2866037,"visible":true,"origin":"","legend":"\u003cp\u003eThe research framework.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a14fba6182c74552049c5238.png"},{"id":96050024,"identity":"4b470e79-f4b3-4342-9ab5-9fe2f83b5802","added_by":"auto","created_at":"2025-11-17 06:36:06","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":106844,"visible":true,"origin":"","legend":"\u003cp\u003eDouble-branch classification network architecture.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/6970c88b06afd6a4102d19df.png"},{"id":96247098,"identity":"867b8279-bbc5-4313-a33e-c159c30c98d6","added_by":"auto","created_at":"2025-11-19 07:27:09","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":527572,"visible":true,"origin":"","legend":"\u003cp\u003eFramework of SegFormer.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/84a3792e594e7166faf3db52.png"},{"id":96050057,"identity":"1d2f8c5f-7a88-4c6e-9fb7-0db125af1b43","added_by":"auto","created_at":"2025-11-17 06:36:10","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":391065,"visible":true,"origin":"","legend":"\u003cp\u003eGrad-CAM image samples created by the classification model.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/23ec6095b7045ddde77f247b.png"},{"id":96363498,"identity":"850a8bc3-363d-489d-947f-6ad58b6064b5","added_by":"auto","created_at":"2025-11-20 10:07:10","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":388605,"visible":true,"origin":"","legend":"\u003cp\u003eTypes and definitions of architectural components/\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/efe5a7fa0b3e738873ef1c05.png"},{"id":96050029,"identity":"805f2876-8dc8-4a1f-aeb5-39741e19263b","added_by":"auto","created_at":"2025-11-17 06:36:07","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":1568310,"visible":true,"origin":"","legend":"\u003cp\u003eFrequency statistics of attention to architectural components in Grad-CAM heatmaps.\u003c/p\u003e","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/014302cdc8747d0480cf77fa.png"},{"id":96050099,"identity":"4f570342-9e67-4d93-80e8-e9d2720ec751","added_by":"auto","created_at":"2025-11-17 06:36:12","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":1042429,"visible":true,"origin":"","legend":"\u003cp\u003eFrequency statistics of attention to roof shapes in Grad-CAM heatmaps.\u003c/p\u003e","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/a4045bb4bc428464c4089a23.png"},{"id":96050025,"identity":"7e3c1803-f6b3-4bf2-abe7-9bdbab725d33","added_by":"auto","created_at":"2025-11-17 06:36:07","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":69907,"visible":true,"origin":"","legend":"\u003cp\u003eStatistics of Attention Frequencies for Architectural Landscape Elements in Grad-CAM Heatmaps. a(Sky); b(Road); c(Mountain); d(Vegetation); e(Water); f(Stone steps); g(Stone balustrade); h(Well); i(Bridge); j(Gate pier); k(Vat); l(Stone); m(Censer); n(Stone mill).\u003c/p\u003e","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/893f3b72d70958747b8c9eae.png"},{"id":96050113,"identity":"81b4c4c5-b221-4885-a565-56b0915f5c73","added_by":"auto","created_at":"2025-11-17 06:36:13","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":4924685,"visible":true,"origin":"","legend":"\u003cp\u003eT-SNE clustering results.\u003c/p\u003e","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/61baf3bfd743fca4bbf001b5.png"},{"id":96369070,"identity":"a89c2115-0282-4ac0-afdd-bd1d20a4b628","added_by":"auto","created_at":"2025-11-20 10:19:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":24260720,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/0b7a96ce-019d-4c25-94ec-b2164f72397a.pdf"},{"id":96050021,"identity":"85eda3f5-4e48-4e0c-8b89-9c59c86eae0b","added_by":"auto","created_at":"2025-11-17 06:36:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":36148,"visible":true,"origin":"","legend":"","description":"","filename":"AppendixA.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/8c90fa5832d78e08e35f1a9f.docx"},{"id":96050038,"identity":"2f374b47-7ce3-4689-801b-5c2a2ea27d25","added_by":"auto","created_at":"2025-11-17 06:36:08","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1522100,"visible":true,"origin":"","legend":"","description":"","filename":"Table1and6.docx","url":"https://assets-eu.researchsquare.com/files/rs-7971684/v1/7ebc2d7b029a2ec179018275.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Identifying and Interpreting Traditional Architectural Style Characteristics Based on Deep Learning","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAs an important material carrier of regional culture, traditional architectural landscape is a material civilization shaped by natural environment, social-economic foundation, folk culture, and other factors in the long-term historical evolution [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Chinese traditional settlements carry a long history and regional culture. However, with the impact of rapid urbanization, traditional settlements are facing a crisis of survival. On the one hand, standardized construction processes have led to the gradual disappearance of regional characteristics of traditional architectures such as horsehead walls. On the other hand, urban expansion has further damaged the regional characteristic of architectural landscapes. As emphasized by Wu [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e], the cultural value of traditional settlements lies in the complex relationship between the architecture and natural environment, which urgently requires further research.\u003c/p\u003e\u003cp\u003eIn recent years, there have been numerous studies on traditional settlement landscapes, such as landscape gene recognition, spatial form quantification, and landscape feature extraction. In terms of landscape gene recognition, Liu [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] took Chinese traditional settlements as research objects, and proposed the “landscape information chain” theory and “cell-chain-shape” analysis model to study traditional settlement landscape gene mining, map construction, and regional division. Hu et al. [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e] constructed a multidimensional traditional settlement landscape genes identification system from spatial form, material elements, and non-material elements, which provided the systematic analysis framework for settlement landscape research. Duan et al. [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] took traditional settlement buildings in Jiangxi and Anhui provinces as examples to construct the cultural landscape factor system from building types, spatial configuration, and natural environment. They summarized the cultural landscape characteristics and differences in this region. For spatial form quantification, Ren et al. [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] used GIS technology and landscape pattern index to analyze the layout characteristics of rural settlements using remote sensing images. Wang [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e] integrated spatial syntax theory and structural linguistic methods to explore the systematic nature of traditional settlement space construction through a comparative study of the construction mechanisms of Han and Zhuang ethnic groups in northern Guangxi province. Regarding landscape feature extraction, Li [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] took the southern region in Shaanxi as a case study and explored a method for identifying and extracting the characteristics and compositional patterns of traditional settlement landscapes in multicultural areas from the settlement to individual buildings. Xing et al. [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] proposed a landscape heritage information extraction and 3D model based on sketches and text prompts, achieving 3D landscape reconstruction. Lu [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] proposed a landscape feature extraction model, which evaluates various natural landscape scene data and balances the accuracy and efficiency requirements of remote sensing landscape classification. These studies have promoted quantitative research on traditional settlement landscapes from different regions and levels, providing academic support for understanding landscape characteristics of traditional settlements.\u003c/p\u003e\u003cp\u003eThe rapid development of deep learning (DL) technology has provided a chance for traditional building recognition. The DL methods for traditional building recognition are divided into image classification, semantic segmentation, and object detection. For image classification, Mathias et al. [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] used building facade images and street images to identify Flemish Renaissance style, Ottoman style, and neoclassical style buildings, laying the foundation for the automated recognition. Han et al. [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] used a convolutional neural network (CNN) to classify traditional architectural styles using building facade images. Tan et al. [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] proposed an automatic classification method for traditional village heritage value elements based on the DL method, taking traditional villages in Hubei Province as the study case. In terms of semantic segmentation, Dai et al. [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] proposed a facade segmentation model based on residential buildings, which accurately identified the components of building facades using street view images. Haznedar et al. [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] used PointNet to segment the architectural elements of Türkiye's Gaziantep heritage, and summarized the architectural style differences in different regions of Türkiye. Sun et al. [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] proposed a weakly supervised semantic segmentation method for ancient architecture based on multi-scale adaptive fusion and spectral clustering, and validated it on the Chinese religious famous mountain ancient architecture dataset and Baroque architecture dataset. For object detection, Xiong et al. [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e] proposed the detection model for traditional Hakka walled houses in China based on ResNet50 and YOLOV2, which achieved excellent performance. Du and Wang [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e] used Bi-YOLO network to identify building components of traditional residences in the southeastern region of Hubei Province. In addition, the technology of feature visualization has improved the interpretability of the research. For example, Obeso et al. [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] classified the architectural styles using digital photos of Mexican cultural heritage, and demonstrated the representative architectural features that the model focuses on through Class Activation Mapping (CAM). Sun et al. [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] analyzed the correlation between architectural style and construction years in Amsterdam, and presented the style evolution pattern using t-distribution random neighbor embedding (T-SNE) clustering method.\u003c/p\u003e\u003cp\u003eWith the popularity of the Internet, many online architectural landscape (OAL) images released by the public have become an important data source for studying traditional architecture. These images are numerous and wide-ranging, featuring both architectural details and architectural environments. Most of them convey representative characteristics of the architectural style. The application of OAL images has also made significant progress. Zu et al. [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] used DL methods to reveal representative features of remote rural buildings from online building photos of Tibetan and Qiang areas, demonstrating the potential of OAL images in analyzing building and environmental features. Two main limitations exist in extracting architectural features based on OAL images. On the one hand, OAL images contain many irrelevant contents to buildings, such as people, sky, and trees, which interfere with the recognition of architectural features; on the other hand, OAL images focus on the facade features of architectures, lacking roof information of the architectures. Fortunately, the top-down perspective provided by remote sensing images can supplement roof and surrounding environmental information [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTherefore, this study integrated RS, DEM and OAL images to interpret environmental characteristics, architectural features and architectural landscape characteristics of the traditional architectures in terms of architectural styles. The RS and DEM data reveal the natural environment that influences the architectural landscape characteristics. RS and OAL images are used to identify the traditional architectural features and reveal the relationship between these features and the overall architectural landscape. The contributions of this study are as follows: (1) Multi-source data (i.e., RS, DEM and OAL data) provides data support and analytic dimension for the interpretation of traditional architectural style features; (2) This paper innovatively proposes to interpret the traditional architecture characteristics from the perspective of architectural styles; (3) The hierarchical framework “architectural environment characteristics-architectural features།architectural landscape characteristics” of Chinese traditional architecture is constructed,, offering a systematic approach for traditional architectural heritage studies.\u003c/p\u003e\n\u003ch3\u003eData Collection and Processing\u003c/h3\u003e\n\u003cp\u003eThe traditional buildings used in this paper are selected from the list of villages in the Digital Museum of Traditional Chinese Settlements, and the architectural styles of the buildings in the villages are clearly documented. The location of the traditional villages used in the paper is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. The list of the traditional villages that covers the six architectural styles is given in Appendix A.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e(1) DEM data: DEM data for each settlement were obtained with a resolution of 12.5m based on the list. These data were used to analyze the topographic and geomorphological features of different architectural styles.\u003c/p\u003e\u003cp\u003e(2) RS image: According to the list, RS images were collected using Tuxin Earth GIS tools with a spatial resolution of 0.4 ~ 0.6m. During the collection process, the criteria of complete preservation and no large-scale occlusion were followed.\u003c/p\u003e\u003cp\u003e(3) OAL image: In the collection of OAL images, a multi-level image acquisition strategy was adopted: firstly, the county name and settlement name were used as keywords for accurate retrieval on search engines such as Google and Baidu, and it ensures the searched images were concentrated on the target settlement. Then, an extended search using representative architectural styles as keywords was added, effectively improving the representativeness of OAL images. Finally, data cleaning was performed on the collected images. The cleaning standards are given as follows: 1) Retain images that show the complete architectures and surrounding environment; 2) Delete images that lack architectural landscape elements. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the RS and OAL images with different architectural styles and gives detailed descriptions.\u003c/p\u003e\u003cp\u003eIn this paper, the dataset of traditional architectures was constructed. It consists of two parts: (1) The architectural style classification dataset includes RS and OAL images. The images were matched each other in each village among different architectural styles. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows the number of RS and OAL images with each architectural style. (2) Semantic segmentation dataset of architectural landscapes only includes OAL images with each architectural style.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eStatistics of CTS image\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStyle\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of CTS\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNumber of RS images\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNumber of OAL images\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHui\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChuan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo facilitate the segmentation of architectural landscape elements, the content of OAL images is divided into two classes, that is, natural and artificial landscape elements. A total of 14 landscape element categories are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. Natural landscape elements include the sky, mountains, vegetation, and water. The sky is the background of the image, and its light changes have a significant impact on visual perception. It occupies a large region in OAL images and is more likely to affect the identification of architectural style. Mountains and water are important natural elements that influence the spatial distribution of traditional settlements [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Mountains and water jointly shape the classic settlement pattern of “nestled against mountains and beside rivers” in space. Vegetation is one of the most common landscapes in settlements and architectures. Artificial landscape elements include roads, stone steps, stone balustrades, bridges, gate piers, censers, vats, stone mills, and wells. Gate piers are located on both sides of residential gates, and their morphological characteristics carry ritual information. Censers are also symbolic of architectural etiquette. Production and living utensils, such as vats, stone mills, and wells, meet daily needs and become unique landscapes in each architectural style.\u003c/p\u003e"},{"header":"Methods","content":"\u003ch2\u003eResearch Framework\u003c/h2\u003e\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the overall workflow of this study includes four main steps: data collection, analysis of architectural environment characteristics, interpretation of architectural features and interpretation of architectural landscape features.\u003c/p\u003e\u003cp\u003e1. Data collection: Based on the list of traditional settlements, the RS, DEM, and OAL images of different architectural styles were collected.\u003c/p\u003e\u003cp\u003e2. Analysis of architectural environment characteristics: Using RS and DEM data, the natural environment characteristics of the architecture for six architectural styles were analyzed.\u003c/p\u003e\u003cp\u003e3. Interpretation of architectural features: Based on RS and OAL images, a double-branch network model was employed for architectural style classification. The heatmap generated by the Grad-CAM method is used to visualize the image regions on which the model focuses on. By counting the frequency and quantity of the highlighted regions, the characteristics of the architectural styles can be clearly presented, such as architectural components, roof types, architectural colors, and architectural materials.\u003c/p\u003e\u003cp\u003e4. Interpretation of architectural landscape features: Based on the OAL images, a semantic segmentation model was employed to divide the images into 14 landscape element categories, which include natural and artificial landscape elements. For each architectural style, firstly, the semantic segmentation results are superimposed with the heatmap to obtain the landscape elements that the model is most concerned about. Then, the six elements with the highest frequency of attention are regarded as representative landscape elements of this architectural style. Finally, T-SNE analysis was conducted on the representative landscape elements, and the landscape correlation characteristics of each architectural style were analyzed.\u003c/p\u003e\u003ch3\u003eDeep Learning Models\u003c/h3\u003e\u003ch2\u003eDouble-branch Classification Network\u003c/h2\u003e\u003cp\u003eBased on the OAL and RS images of traditional architectures, a double-branch network was designed for the classification of the architectural styles. As shown in Figure.4, the proposed model consists of feature extraction and feature fusion. In the feature extraction, a DenseNet backbone was employed for both branches to facilitate robust feature learning [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Specifically, the RS branch specializes in learning the architectural roof morphology and overall layout features, while the OAL branch focuses on the landscapes of the architectures, door and window styles, the texture and materials of the walls. In the feature fusion, the CBAM mechanism was introduced. This mechanism utilizes dual weighting of channel and spatial attention to automatically focus on the most discriminative building features [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. In addition, the module also incorporated a feature reuse mechanism to achieve cross branch fusion of high-level features from the OAL branch. For a detailed introduction of the model, please refer to the paper of Zhang et al. [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e\u003ch3\u003eSemantic Segmentation Network\u003c/h3\u003e\u003cp\u003eTo perform architectural landscape semantic segmentation on OAL images, SegFormer was employed. SegFormer is an advanced semantic segmentation model based on the Transformer architecture [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e], which adopts a combination architecture of layered Transformer encoder (MiT) and lightweight multilayer perceptron (MLP) decoder. The input image is transformed into a multi-scale feature representation through the overlapping block embedding, which can simultaneously capture the detailed features of the building and the overall spatial layout relationships to achieve accurate segmentation [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. The network architecture of the model is shown in Figure. 5.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e shows the framework of the SegFormer. The encoder part adopts a hierarchical design, including multiple stages of Transformer modules. Each stage constructs a multi-scale feature pyramid from high-resolution details to low resolution semantics by gradually reducing sequence length and expanding channel dimensions. For an image with the input size of \u003cem\u003eH\u003c/em\u003e×\u003cem\u003eW\u003c/em\u003e×3, SegFormer first uses the overlap patch embeddings technique to divide the image into multiple patches of the size 4×4. Pre-trained CNN is used to extract features from each block and convert them into high-dimensional vectors. Subsequently, these patches are input to a hierarchical Transformer encoder to generate multi-level features with original image resolutions of {1/4, 1/8, 1/16, 1/32}. Efficient self-attention is used in the stage to effectively reduce computational complexity, and a 3×3 convolution operation is introduced to the Hybrid Feedforward Network (Mix-FFN). Mix-FFN can extract global contextual information and preserve local spatial details, significantly enhancing the ability to represent complex structures and subtle features of architectural images. The overlap patch merging technique served as a complementary image processing step, including dividing overlapping blocks, extracting feature, averaging overlapping regions, and merging block to restore overlapping image blocks to a complete image.\u003c/p\u003e\u003cp\u003eThe decoder abandons the complex attention mechanism or upsampling module in traditional structures and only uses a lightweight multi-layer perceptron (MLP) to fuse the multi-scale features that output by different levels of encoders. The specific process of the full MLP decoder includes four steps: first, the multi-level features that output by the MiT encoder are unified into channel dimensions through the MLP layer; Next, upsample each feature to a resolution of 1/4 of the original image and concatenate them; Subsequently, the concatenated features are fused through the MLP layer; Finally, after another layer of MLP, the segmentation mask with a resolution of H/4×W/4×N\u003csub\u003ecls\u003c/sub\u003e (N\u003csub\u003ecls\u003c/sub\u003e is the number of categories) is output. It is worth noting that the overlapping block strategy in SegFormer makes it particularly suitable for handling complex contours of traditional buildings, such as horsehead walls, effectively avoiding the common edge aliasing problem in CNN [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Meanwhile, its powerful self-attention mechanism can establish long-range dependency relationships, accurately associate dispersed building features, and further enhance the accuracy and reliability of traditional building semantic segmentation.\u003c/p\u003e\u003ch2\u003eGrad-CAM Method\u003c/h2\u003e\u003cp\u003eGrad-CAM method is used to visualize the classification and segmentation networks [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. It can highlight the importance of each position in the feature map of a given category by generating heatmaps. First, the last convolutional layer in the feature extraction module is selected, and the gradient of the target category with respect to feature maps of this layer is calculated. Then, the global average of the gradients on each channel is computed for the \u003cem\u003ek\u003c/em\u003eth feature map to obtain the weight of the feature map. Finally, the feature map \u003cem\u003eA\u003c/em\u003e\u003csup\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sup\u003e is accumulated according to the weight, and the \u003cem\u003eReLU\u003c/em\u003e function is used to filter out negative values, and the heatmap of category \u003cem\u003ec\u003c/em\u003e is obtained. The formula is computed as follows:\u003c/p\u003e\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\alpha _{k}^{c}=\\frac{1}{Z}\\sum\\limits_{i} {\\sum\\limits_{j} {\\frac{{\\partial {y^c}}}{{\\partial A_{{ij}}^{k}}}} }$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$L_{{Grad-CAM{\\text{ }}}}^{c}=ReLU(\\sum\\limits_{k} {\\alpha _{k}^{c}} {A^k})$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003cp\u003eIn the formula, \u003cem\u003eAk ij\u003c/em\u003e represents the pixel value at row \u003cem\u003ei\u003c/em\u003e and column \u003cem\u003ej\u003c/em\u003e in the \u003cem\u003ek\u003c/em\u003eth feature map, \u003cem\u003eZ\u003c/em\u003e is the number of pixels in the feature map, \u003cem\u003ey\u003c/em\u003e\u003csup\u003e\u003cem\u003ec\u003c/em\u003e\u003c/sup\u003e is the category score of the \u003cem\u003ec\u003c/em\u003eth class; \u003cem\u003eαc k\u003c/em\u003e is the weight of the \u003cem\u003ek\u003c/em\u003eth feature map corresponding to the \u003cem\u003ec\u003c/em\u003eth class. \u003cem\u003eA\u003c/em\u003e\u003csup\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sup\u003e is the \u003cem\u003ek\u003c/em\u003eth feature map of the convolutional layer, and \u003cem\u003eLc Grad-CAM\u003c/em\u003e is the heatmap result of category \u003cem\u003ec\u003c/em\u003e.\u003c/p\u003e\u003ch3\u003eT-SNE Clustering Method\u003c/h3\u003e\u003cp\u003eT-SNE method is employed to analyze the clustering results of semantic segmentation which is produced based on the OAL images. It is a dimensionality reduction method that reduces the high-dimensional feature of each sample to a low-dimensional space to visualize feature changes [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Samples with similar features are aggregated, while those with different features are separated. The formula is computed as follows:\u003c/p\u003e\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$${p_{j|i}}=\\frac{{\\exp ( - ||{x_i} - {x_j}|{|^2}/2{\\sigma _i}^{2})}}{{\\sum\\limits_{{}} {_{{k \\ne i}}\\exp } ( - ||{x_i} - {x_k}|{|^2}/2{\\sigma _i}^{2})}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$${p_{ij}}=\\frac{{{p_{j|i}}+{p_{i|j}}}}{{2n}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003cp\u003eIn the formula, \u003cem\u003ex\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003ex\u003c/em\u003e\u003csub\u003e\u003cem\u003ej\u003c/em\u003e\u003c/sub\u003e represent two high-dimensional data points. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:\\)\u003c/span\u003e\u003c/span\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e is Gaussian kernel width of \u003cem\u003ex\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e, which is a parameter used to adjust local range of data points, \u003cem\u003ep\u003c/em\u003e\u003csub\u003e\u003cem\u003ei|j\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003ep\u003c/em\u003e\u003csub\u003e\u003cem\u003ej|i\u003c/em\u003e\u003c/sub\u003e represent the conditional probability of similarity of data points in high-dimensional space, \u003cem\u003ep\u003c/em\u003e\u003csub\u003e\u003cem\u003eij\u003c/em\u003e\u003c/sub\u003e is the joint probability, and \u003cem\u003en\u003c/em\u003e is the number of data points. \u003cem\u003ek\u003c/em\u003e is a summation index that iterates over all data points \u003cem\u003ex\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e except for \u003cem\u003ex\u003c/em\u003e\u003csub\u003e\u003cem\u003ei.\u003c/em\u003e\u003c/sub\u003e\u003c/p\u003e\u003cdiv id=\"Equ5\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ5\" name=\"EquationSource\"\u003e\n$${q_{ij}}=\\frac{{{{(1+||{y_i} - {y_j}|{|^2})}^{ - 1}}}}{{\\sum\\nolimits_{{k \\ne l}} {{{(1+||{y_k} - {y_l}|{|^2})}^{ - 1}}} }}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\u003c/div\u003e\u003cp\u003eIn the formula, \u003cem\u003ey\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003ey\u003c/em\u003e\u003csub\u003e\u003cem\u003ej\u003c/em\u003e\u003c/sub\u003e are the mapped low dimensional data, and \u003cem\u003eq\u003c/em\u003e\u003csub\u003e\u003cem\u003eij\u003c/em\u003e\u003c/sub\u003e is the conditional probability that the data is similar in the low dimensional space. \u003cem\u003ek\u003c/em\u003e and \u003cem\u003el\u003c/em\u003e are summation indices that iterate over all pairs of distinct low-dimensional points \u003cem\u003ey\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003ey\u003c/em\u003e\u003csub\u003e\u003cem\u003el\u003c/em\u003e\u003c/sub\u003e.\u003c/p\u003e\u003cdiv id=\"Equ6\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ6\" name=\"EquationSource\"\u003e\n$$C=KL(P||Q)=\\sum\\limits_{i} {\\sum\\limits_{j} {{p_{ij}}} } \\log \\frac{{{p_{ij}}}}{{{q_{ij}}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e6\u003c/div\u003e\u003c/div\u003e\u003cp\u003eIn the formula, \u003cem\u003eP\u003c/em\u003e and Q represent the similarity distribution between data points in high-dimensional and low dimensional space, respectively. \u003cem\u003eKL\u003c/em\u003e is Kullback Leibler divergence, a tool used to measure the difference between two distributions. \u003cem\u003eC\u003c/em\u003e represents the degree of mismatch between \u003cem\u003ep\u003c/em\u003e\u003csub\u003e\u003cem\u003eij\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eq\u003c/em\u003e\u003csub\u003e\u003cem\u003eij\u003c/em\u003e\u003c/sub\u003e.\u003c/p\u003e"},{"header":"Results and Analysis","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eExperimental Preparation\u003c/h2\u003e\u003cp\u003eThe RS and OAL images in the traditional architecture dataset were processed as follows: All images were cropped to 256\u0026times;256 pixels and divided into training, validation, and test sets at a ratio of 6:2:2. At the same time, data augmentation were used to expand the dataset, such as rotation, mirroring and color enhancement. To ensure the fairness and accuracy of the experiment, the same experimental environment and initial training parameters were used, which is shown in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eOverview of experimental environment and parameters\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eExperimental environment\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCPU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIntel(R) Xeon(R) CPU E5-2673 v4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGPU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNVIDIA GeForce RTX 3060 Ti, 12 GB memory\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCUDA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11.8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePyTorch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.0.0\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePython\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3.8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eParameters of double-branch classification network\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eResize\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSize= (256,256)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBatchsize\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEpoch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLearning rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1e-2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOptimizer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSGD\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLoss function\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCrossEntropyLoss\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eParameters of semantic segmentation network\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eResize\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSize= (256,256)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBatchsize\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEpoch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e100\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLearning rate\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1e-4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOptimizer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAdamW\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWeight decay\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1e-4\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eModel Evaluation Metrics\u003c/h2\u003e\u003cp\u003eTo evaluate the performance of the architectural style classification experiments and architectural landscape segmentation experiments, some metrics are introduced here. In the classification experiments, accuracy, precision, recall, and F1-score were employed to evaluate the performance of the double-branch classification network. They are computed as follows:\u003cdiv id=\"Equ7\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ7\" name=\"EquationSource\"\u003e\n$$Accuracy=\\frac{{TP}}{{TP+FP+FN+TN}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e7\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ8\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ8\" name=\"EquationSource\"\u003e\n$$Precision=\\frac{{TP}}{{TP+FP}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e8\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ9\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ9\" name=\"EquationSource\"\u003e\n$$Recall=\\frac{{TP}}{{TP+FN}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e9\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ10\" class=\"Equation\"\u003e\u003cdiv class=\"EquationNumber\"\u003e10\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eIn landscape segmentation experiments, category average pixel accuracy (mPA), average precision (mPrecision), and average intersection to union ratio (mIoU) are used as evaluation metrics to evaluate the performance of SegFormer. They are computed as follows:\u003cdiv id=\"Equ11\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ11\" name=\"EquationSource\"\u003e\n$$mPA=\\frac{1}{k}\\sum\\nolimits_{{i=1}}^{k} {\\frac{{T{P_i}}}{{T{P_i}+F{N_i}}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e11\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ12\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ12\" name=\"EquationSource\"\u003e\n$$mPrecision=\\frac{1}{k}\\sum\\nolimits_{{i=1}}^{k} {\\frac{{T{P_i}}}{{T{P_i}+F{P_i}}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e12\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ13\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ13\" name=\"EquationSource\"\u003e\n$$mIoU=\\frac{1}{k}\\sum\\nolimits_{{i=1}}^{k} {\\frac{{T{P_i}}}{{T{P_i}+F{P_i}+F{N_i}}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e13\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eIn the above formulas, \u003cem\u003eTP\u003c/em\u003e, \u003cem\u003eFP\u003c/em\u003e, \u003cem\u003eTN\u003c/em\u003e and \u003cem\u003eFN\u003c/em\u003e represent true positive, false positive, true negative, and false negative respectively. \u003cem\u003ek\u003c/em\u003e represents target categories, and \u003cem\u003ei\u003c/em\u003e denotes the object index.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eAnalysis of Architectural Environment Characteristics\u003c/h2\u003e\u003cp\u003eAs the material basis for the formation of traditional settlement landscapes, the natural environment directly affects the spatial layout and morphological characteristics of the settlements. Specifically, terrain is the most crucial element, directly influencing the distribution pattern of water and vegetation within the settlements, as well as the formation of soil [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Water is a necessary condition for human survival, which affects the location and layout of settlements. The selection of vegetation in settlement landscapes is not only based on natural factors, but also on the factors of production and daily life [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eStatistics of the number of settlement landform types\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eArchitectural style\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNumber of settlements\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ePlains\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHills\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMountains\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePlateaus\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHui\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChuan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e18\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eTo have a more comprehensive understanding of the environmental characteristics of architectures for each architectural style, the natural environment, including terrain, water and vegetation, was analyzed using RS and DEM data. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, the types of terrain are divided into plain, hill, mountain, and plateau. Specifically, (1) the architectures of Su are mainly distributed in plain areas, which have flat terrain, and the settlements are distributed in high-density blocks. The settlements in this area are either located within river networks or distributed along riverbanks. Artificial vegetation and herbs along the rivers are very common in this area. The garden style layout of the architectures shows the beauty of the integration of architectures and natural landscape. (2) The architectures of Hui are mostly distributed in hills with flat terrain, which is easy to carry out construction activities and can avoid the risk of floods. These settlements are near rivers, and most of the rivers are mountain streams and small rivers. The vegetation coverage in this area is relatively high, and mostly a mixture of arbors and shrubs, forming a continuous landscape with the surrounding forest land. (3) The architectures of Jing are mainly located in hills and mountains. The hilly area has a gentle terrain, and the water is mostly seasonal rivers and artificially excavated ditches. The vegetation is mainly composed of arbor. In the mountainous areas, the height of the mountain is relatively low. The layout of the architectures is adjusted according to the slope and direction. The forest coverage rate is higher in mountainous areas. (4) The architectures of Min are mostly located in wooded mountains or river valleys surrounded by mountains. There are many perennial rivers in the river valley area, and the vegetation in the mountains is mostly broad-leaved forests. The rivers and closed terrain together form a natural barrier for family settlement. (5) The architectures of Chuan are mostly located in mountainous areas with steep terrain. The cantilevered structure of the stilt house adapts to mountain slopes, which is convenient for ventilation. There are numerous rivers, and the surrounding vegetation is mainly bamboo forests and broad-leaved shrubs. (6) The architectures of Jin are mainly distributed in mountains and plateaus, and the terrain mainly consists of steep mountains and the Loess Plateau. The buildings in mountainous areas adopt deep eaves and thick walls, which can withstand severe cold and temperature changes. The water is mainly mountain streams and seasonal rivers, and vegetation is scattered shrubs with a relatively low coverage. The terrain of the Loess Plateau is undulating, and architectures here make full use of loess resources. The cave dwellings are typical architectures. The water is dominated by seasonal rivers, and the vegetation coverage is low.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eInterpretation of Architectural Features\u003c/h2\u003e\u003cp\u003eIn this chapter, the architectural styles are firstly classified to achieve intelligent recognition of architectural styles. Then, using Grad-CAM, representative features of each architectural style are obtained. Finally, architectural features are interpreted for different architectural styles.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eClassification of Architectural Styles\u003c/h2\u003e\u003cp\u003eIn order to select the most suitable network to class the architectural style, ResNet152, MobileNetV2, EfficientNet-B0 were used to replace the network in the feature extraction modules for comparative experiments, and these models were named RDB-Net, MDB-Net, EDB-Net, respectively [\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. In the experiment, all models used the same training and testing datasets for fairness. Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e lists the evaluation metrics and their average precision values of the comparative models. The results show that the proposed model achieved the highest average value of three indicators, namely, recall (89.73%), accuracy (90.37%), and F1-score (89.60%). The proposed model has been identified as the most effective model for identifying architectural styles.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eEvaluation indicators for classification models\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"9\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eStyle\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSu\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eHui\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eJing\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eChuan\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eJin\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eAverage\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eMDB-Net\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e89.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e94.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e68.66\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e89.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e96.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e81.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e86.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e88.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e91.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e76.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e92.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e92.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e78.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e86.59\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eF1-score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e88.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e92.65\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e72.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e90.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e80.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e86.62\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eRDB-Net\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e85.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e71.64\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e91.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e96.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e75.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e84.76\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e82.61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e76.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e82.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e81.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e91.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e98.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e85.47\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eF1-score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e83.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e81.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e76.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e85.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e85.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e84.68\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eEDB-Net\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e80.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e95.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e79.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e93.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e74.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e85.24\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e82.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e68.83\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e96.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e96.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e79.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e86.10\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eF1-score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e73.61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e92.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e95.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e76.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e85.40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eOurs\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRecall\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e92.54\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e98.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e89.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e98.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e71.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e\u003cb\u003e89.73\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e89.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e77.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e95.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e90.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e95.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e\u003cb\u003e90.37\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eF1-score\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e93.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e82.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e92.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e81.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e\u003cp\u003e\u003cb\u003e89.60\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBased on the classification model, the highlight areas in the Grad-CAM heatmap of OAL images can visualize representative features of the architectural style, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. The brighter areas in the heatmap represent the regions that the model focuses on, and the redder the color, the higher the attention it receives. The model focuses on the architecture itself, rather than other backgrounds, which verifies the reliability of the classification model.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eInterpretation of Architectural Features for Each Architectural Styles.\u003c/b\u003e\u003c/p\u003e\u003cp\u003eAccording to Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e, the model focuses on both the overall structure and the detailed elements of the architectures. To explore the key characteristics that affect the classification of architectural style, architectural components, roof types, architectural colors and architectural materials are analyzed in this chapter. The features with the frequency of less than 10 times are not considered as the representative features, and thus it will not be discussed here.\u003c/p\u003e\u003cp\u003e1. Architectural components\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents the types and definitions of architectural components. According to the Grad-CAM heatmaps, the attention frequency of different architectural components for each architectural style is counted, which is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e. As the important component of architectures, the roof has the highest attention frequency in all architectural styles. It reflects the crucial role of the roof in shaping architectural styles. The roofs of different architectural styles have diverse forms and characteristics, which will be discussed in the next subsection.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eExcept for the roof, wooden windows are the main components in all architectural styles. The horsehead wall is the key component of Hui. Plaques often appear in the Jin and Min, while stone windows often appear in the Su and Hui. Specifically, (1) in the architectures of Su, wooden windows are the components that receive the most attention except for the roof. These windows emphasize decoration, and integrate with surrounding landscapes, such as mountains, water, flowers and trees. As a leaky window in garden architectures, the rich patterns of stone windows make them highly decorative. Columns and railings, as essential elements of garden architectures, pay attention to carving and decoration, and have a high frequency of attention. (2) In the architectures of Hui, horsehead walls are the most recognizable components except for the roof. Their structures can prevent fires. Doors with exquisite brick and wood carvings are important symbols of art and family wealth. (3) In the architectures of Jing, doors are the components that receive the most attention except for the roof. Their shape and scale adhere strict hierarchy and regulations, representing the family status and wealth level. The attention frequency of columns is also relatively high, and their arrangement and size follow strict construction regulations. (4) In the architectures of Min, wooden windows have the most attention except for the roof. The wooden windows have white frames and simple forms of smaller size. The gates of the Tulou are heavy and sturdy, equipped with solid locks and defense facilities. Plaques often display the family name, or auspicious words. (5) In the architectures of Chuan, railings are the components that receive the most attention except for the roof. Numerous railings have been set up in platforms, pavilions and other spaces. The arrangement and size of columns are designed according to the terrain characteristics. The overall height of the structures can be adjusted to adapt to complex terrain. (6) In the architectures of Jin, doors have the most attention except for the roof, and their shape and scale are extremely exquisite, decorating with brick carvings. Plaques are commonly found in temples and ancestral halls. The decoration consists of lanterns and stone carvings. Lanterns are commonly used in festival scenes, while stone carvings are applied to lintels and column foundations.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e2. Roof type\u003c/p\u003e\u003cp\u003eThe types and characteristics of the roofs are shown in Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. The Grad-CAM heatmaps were used to calculate the attention frequency of different roof types for each architectural style, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e.\u003c/p\u003e\u003cp\u003eOverall, hard mountain roofs are widely used in the architectures of Su, Hui, Jing, and Jin. Overhanging gable roofs are used in the architectures of Min and Chuan, and pyramid roofs are used in Su and Chuan. Specifically, (1) in the architectures of Su, the gable-on-hip roofs are primary roof styles, which are commonly used in ancestral halls and temples. The pyramid roofs frequently appear in Jiangnan gardens paired with pavilions and towers, which serve as prominent visual focal point in the landscape. Hard mountain roofs are commonly used in residential buildings to meet the practical needs of fire and wind prevention. (2) In the architectures of Hui, hard mountain roofs are the primary roof types. The gable-on-hip roofs are primarily used in ancestral halls, temples, or similar structures to emphasize their functional importance. (3) In the architectures of Jing, hard mountain roofs are commonly used in residential buildings, characterized by low construction cost. The hip roofs are main roof types of the palaces, symbolizing the supreme imperial power. In addition to hip roofs, gable-on-hip roofs are also found on the palace side hall, temples and some official buildings, serving both aesthetic and hierarchical purposes. (4) In the architectures of Min, overhanging gable roofs are the main roof type. The large overhanging eaves of Tulou not only provide shade for walls and doors but also facilitate effective drainage. (5) The overhanging gable roofs are primary roof types in the architectures of Chuan, which play a role in drainage. Pyramid roofs and gable-on-hip roofs are main roof types for bamboo houses, drum towers, and wind-rain bridges, meeting needs for sightseeing, relaxation, and clan activities. (6) Hard mountain roofs are commonly used in the architectures of Jin, and they can effectively resist the erosion of wind and sand on the building walls and roofs. Flat roofs are common in cave buildings, and they matche the structure of the cave, providing the space for drying crops and storing debris.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e3. Architectural colors\u003c/p\u003e\u003cp\u003eAccording to OAL images, architectural colors of each image were counted, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. It can be seen from the table that main colors of the architectures of Su and Hui are black, white and gray. Golden and vermillion are the common color in the palaces of Jing, while black and reddish brown are main color in the residential buildings. Specifically, (1) in the architectures of Su, walls are off white, and roofs are covered with small black tiles. The door and window columns have two colors, one is black or dark brown, and the other is reddish brown. (2) The architecture of Hui is deeply influenced by Confucian culture, and most of the columns of doors and windows are black or dark brown. (3) In the architectures of Jing, palaces predominantly feature vermilion walls and golden glazed tiles, and the columns of doors and windows are also vermilion. The color conforms to the aesthetic pursuit of solemnity and magnificence in royal culture. The walls of residential buildings are white or tan. Door and window columns are primarily reddish brown or dark brown, with reddish brown being more common. (4) In the architectures of Min, the exterior walls are rammed with local soil, resulting in an earth yellow color. The doors are commonly used reddish brown, which has the meaning of exorcising evil in Hakka culture. (5) In the architectures of Chuan, the regions where the architectures are rich in bamboo and wood resources, and the main color of the architectures is dark brown. (6) In the architectures of Jin, the wall color of Shanxi merchants' architecture is dark gray, with a small amount of earth yellow. The wall color of the cave is mainly earth yellow.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eQuantity statistics of architectural colors based on OAL images\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStyle\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWall color\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eColor of door and window columns\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eRoof color\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCount\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOff white\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e205\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOff white\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e129\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eHui\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOff white\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e309\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOff white\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eJing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePalace\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eVermilion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eVermilion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eGolden\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eQuadrangle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWhite or tan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e157\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWhite or tan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eTulou\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEarth yellow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e329\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOff white and reddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBlack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eChuan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStilted building\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e270\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDrum tower and Wind-rain bridge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown and white\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e55\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBamboo building\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eJin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eShanxi Merchants' Architecture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown and reddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e164\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEarth yellow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e20\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCave dwelling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown and reddish brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark gray\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e23\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eEarth yellow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDark brown\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e127\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e4. Architectural materials\u003c/p\u003e\u003cp\u003eBased on OAL images, architectural materials used for each style were statistically analyzed, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab8\" class=\"InternalRef\"\u003e8\u003c/span\u003e. Southern architectures primarily use bamboo, wood, and rammed earth. Local materials are used for humid climates and complex topography. Northern architectures utilize bricks, stones, tiles, and loess to adapt to arid conditions.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab8\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 8\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eStatistics of architectural materials based on OAL images\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStyle\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWall material\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eRoof material\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eThe material of door and window columns\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCount\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSu\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHui\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eJing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePalace\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eGlazed tiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood and white marble\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eQuadrangle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e236\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTulou\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRammed earth\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e334\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eChuan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStilted building\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles or thatch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e270\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDrum tower\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood or earth-stone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWind-rain bridge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood or stone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e41\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBamboo building\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBamboo-wood mixture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eThatch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBamboo-wood mixture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eJin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eShanxi Merchants' Architecture\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e184\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eCave dwelling\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLoess\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWood and thatch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e110\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLoess, stone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTiles, stone and brick\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWood\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eSpecifically, (1) the architectures of Su and Hui are consistent in materials. The walls are made of solid bricks, the roofs are made of small blue tiles, which play a rainproof role. The columns of door and window are made of wood, and they are good at making wood carvings for decoration. (2) In the architectures of Jing, palace roofs are covered with glazed tiles, and door and window columns are made of white marble stone, highlighting the elegance and magnificence. The roofs of residential buildings are paved with ordinary tiles. (3) In the architectures of Min, the wall materials take rammed earth as the core materials, and its components are clay, sand and lime, mixed with bamboo to enhance the structural strength. It has the functions of heat preservation, heat insulation and defense. (4) In the architectures of Chuan, stilted buildings use wood as the material of the columns of walls and door columns, and the roof uses small green tiles or straw. The roofs of bamboo tower of Dai are covered with thatch and made of a mixture of bamboo and wood materials, which can resist the erosion of rainwater and insects. (5) In the architectures of Jin, the walls of Shanxi merchants' architectures are made of bricks, offering high defensive capability and effective resistance to cold and sandstorm. The roofs are covered with tiles to accelerate drainage. Small sized cave dwellings use loess as the wall material to achieve the purpose of warm in winter and cool in summer. In the larger sized courtyards, the roofs are made of bricks and tiles with stronger load-bearing capacity, and the walls are made of a mixed structure of loess and bricks, which enhances durability through material composites.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eInterpretation of Architectural Landscape Features\u003c/h2\u003e\u003cp\u003eTo find the most representative architectural landscape features for each architectural style, the semantic segmentation experiment in OAL images is firstly carried out. Combined with the Grad-CAM heatmaps, the representative architectural landscape elements are selected. Then, the semantic clustering results are analyzed with T-SNE model to explore the semantic association and potential confusion reasons for each architectural style.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eSemantic Segmentation of Architectural Landscape\u003c/h2\u003e\u003cp\u003eTo select the most suitable model to segment the architectural landscape elements, several common segmentation models were compared, such as Segmenter, DeepLabV3+, UNet, PSPNet, Swin Transformer and SegFormer. In the experiment, all models used the same training and testing datasets for fairness. Table\u0026nbsp;\u003cspan refid=\"Tab9\" class=\"InternalRef\"\u003e9\u003c/span\u003e lists the evaluation metrics and the average precision of the comparative models for each segmentation elements. The results show that SegFormer achieves the highest performance among the average values of three indicators: pixel accuracy (PA), accuracy (P), and Intersection over Union (IoU). Therefore, the SegFormer was selected as the model to segment the architectural landscape elements on OAL images.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab9\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 9\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eEvaluation indicators of segmentation models\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003eModel\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSegmenter\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eDeepLabV3+\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eUNet\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePSPNet\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eSwin Transformer\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eSegFormer\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003emPA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e68.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e79.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e80.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e82.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e83.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e84.11\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003emPrecision\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e83.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e86.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e87.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e89.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e87.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e89.92\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e\u003cp\u003emIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e61.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e70.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e72.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e74.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e74.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e\u003cb\u003e76.44\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eSky\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e96.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e96.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e97.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e96.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e98.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e97.03\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e93.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e95.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e96.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e95.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e96.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e98.14\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e90.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e92.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e94.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e91.63\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e95.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e95.28\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eBuilding\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e94.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e94.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e94.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e94.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e94.99\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e90.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e93.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e93.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e93.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e95.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e95.22\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e90.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e90.67\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eRoad\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e77.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e86.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e89.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e90.88\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e90.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e85.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e85.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e85.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e89.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e87.75\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e72.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e75.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e76.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e77.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e81.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e80.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eMountain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e92.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e93.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e93.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e93.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e91.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e93.86\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e89.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e92.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e90.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e93.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e91.29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e93.35\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e82.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e87.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e85.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e87.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e84.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e87.98\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eVegetation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e85.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e87.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e87.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e86.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e90.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e90.77\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e89.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e89.95\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e75.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e78.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e79.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e78.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e81.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e82.41\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eWater\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e89.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e94.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e93.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e94.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e95.45\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e95.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e92.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e95.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e93.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e95.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e93.30\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e86.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e86.98\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e90.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e89.33\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eStone steps\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e45.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e61.88\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e66.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e62.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e58.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e67.59\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e83.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e86.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e82.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e88.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e83.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e88.64\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e48.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e56.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e58.52\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e58.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e52.45\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e62.21\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eStone balustrade\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e72.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e80.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e81.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e83.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e85.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e85.68\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e72.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e82.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e78.85\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e76.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e88.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e81.96\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e57.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e68.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e66.54\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e66.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e77.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e72.08\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eWell\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e42.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.75\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e87.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e91.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e94.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e90.61\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e59.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e68.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e91.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e90.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e82.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e91.75\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e50.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e62.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e80.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e83.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e79.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e83.78\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eBridge\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e63.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e78.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e75.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e77.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e81.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e77.73\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e79.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e81.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e86.68\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e86.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e82.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e89.50\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e54.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e66.70\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e67.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e69.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e69.34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e71.23\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eGate pier\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e53.80\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e62.16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e63.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e73.30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e69.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e91.02\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e90.61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e84.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e81.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e85.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e81.08\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e91.55\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e55.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e55.49\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e65.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e59.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e83.97\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eVat\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e44.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e88.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e74.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e91.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e97.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e92.87\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e98.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e91.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e76.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e92.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e75.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e72.55\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e44.36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e81.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e60.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e85.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e74.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e68.73\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eStone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e85.31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e79.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e82.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e83.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e88.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e82.32\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e74.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e84.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e85.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e85.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e87.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e85.79\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e65.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e69.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e72.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e73.18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e78.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e72.44\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eCenser\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7.24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17.39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e27.69\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e35.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e26.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e29.12\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e64.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e87.61\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e82.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e91.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e86.82\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e95.99\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e16.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e26.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e34.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e25.41\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e28.77\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eStone mill\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e75.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e77.28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e91.73\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e78.76\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e88.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e81.65\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e88.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e90.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e96.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e93.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e93.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e93.30\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eIoU\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e68.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e71.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e88.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e74.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e83.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e77.13\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eRepresentative Architectural Landscape of Different Architectural Styles\u003c/h2\u003e\u003cp\u003eBased on the segmentation results of the architectural landscape elements, the attention frequency of the Grad-CAM heatmaps on each architectural style is counted, as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e. The statistical results reflect the significance of different architectural landscape elements in the classification process for each architectural style. These statistics can better reflect the repeatability and stability of the architectural landscape elements in architectural style discrimination. The more frequently it occurs, the greater the contribution of the landscape element. It should be pointed out that elements with the frequency of less than 10 times are not considered as the representative element, and thus they will not be discussed here.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFor each architectural style, the top six architectural landscape elements with the highest frequency are selected as the representative landscape elements from 14 kinds. Sky, road, and vegetation are landscape elements that all architectural styles pay attention to. Water, mountain, stone, well and bridge commonly appear in the southern architectural landscape; Gate pier, vat, center and stone mill commonly appear in the northern architectural landscape. Stone steps and stone balustrade commonly appear in the architectures of Jing. In addition, each architectural style also has some unique landscape elements, which have unique cultural connotation and landscape value.\u003c/p\u003e\u003cp\u003eSpecifically, (1) in the architectures of Su, the representative architectural landscape elements are vegetation, sky, water, stone, road and bridge. Water, bridges, and stones are the core landscape symbols of garden architectures. These elements highlight the garden feature by the integration of architectures and nature. (2) In the architectures of Hui, the representative architectural landscape elements are vegetation, sky, mountains, water, roads and bridges. The combination of mountains and water highlights the layout of settlements. The bridge is an important connection, and stone arch bridges and stone slab bridges are relatively common. The gate pier is located on both sides of the gate, carved with exquisite patterns. Stone steps are commonly found at the entrance of important architectures such as ancestral halls and temples, with a simple and solid structure. (3) In the architectures of Jing, the representative landscape elements are sky, roads, vegetation, stone steps, mountains and stone balustrade. Stone steps and stone balustrade are the key landscape elements. Stone balustrade carved with dragons and phoenixes is widely used in the palaces such as the Forbidden City, which highlights the dignity of the royal family. Stone steps are used in palaces and temples, and their materials, widths, and levels are governed by strict regulations for different architectural grades. The door piers located on both sides of the gate are carved with lions and drums, which symbolize social status. The vat with goldfish and lotus is commonly placed in the courtyard, which adds vitality and elegance to the courtyard. Censer is commonly placed in the temples and palaces, and is used for activities of sacrifice. (4) The representative landscape elements of Min are mountains, vegetation, sky, roads, water and wells. Wells are the key landscape element with regional characteristics. They have met the residents' demand for domestic water under the settlement pattern. (5) The representative landscape elements of Chuan are mountains, sky, vegetation, water, roads and stone balustrade. As the protective and decorative element, stone balustrade is commonly used at the edge of bridges, plank roads and building platforms. Their carving patterns are integrated with the folk culture elements of the Bashu region. (6) In the architectures of Jin, the representative landscape elements are the sky, roads, mountains, vegetation, stone steps and gate piers. The gate piers are often carved with rich patterns. Together with the stone steps, they highlight the architectural rituals and defense needs in the Loess Plateau. The vat not only can store water, but also can adjust the microclimate in the courtyard. The stone mill is a symbol of traditional agricultural culture, and it was once an indispensable production tool in rural life.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eSemantic Clustering Analysis\u003c/h2\u003e\u003cp\u003eThe semantic clustering analysis is carried out with T-SNE model to explore the semantic association and potential confusion reasons of representative architectural landscape elements for each architectural style. It can further understand the cognitive logic of the model to traditional architectural landscapes.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003e shows the semantic clustering results of representative architectural landscapes for each architectural style. Specifically, (1) in the architectures of Su, the semantic clustering of the landscapes presents distinctive garden features. As the unique semantic feature of the Su, vegetation, stone and water together create a natural landscape, which leads to semantic confusion. (2) In the architectures of Hui, laying roads along water is a typical layout of Hui, and the spatial dependence leads to a blurred semantic boundary between them. In addition, the mountains are covered with vegetation, leading to semantic confusion. (3) In the architectures of Jing, the unique semantic connection of landscape elements is reflected in the association of stone balustrade, stone steps and roads. Stone steps are used as the extension of roads, and stone balustrade is defined as the scope of roads. Semantic confusion occurs due to the closeness of their spatial positions. (4) In the architectures of Min, as important life facilities in Tulou buildings, wells are directly connected to internal roads, which leads to semantic confusion. In addition, the mountains are covered with vegetation, leading to semantic connection. (5) In the architectures of Chuan, mountains, vegetation, and water form a close semantic connection. In addition, vegetation around the architectures and the stone railings around the stilt houses intersect in space, and the relationship between architectures and nature creates a semantic confusion. (6) For the architectures of Jin, the semantic association of landscape elements highlights the environmental characteristics of the Loess Plateau. The stone steps are located at the entrance of the building, and the gate piers are adjacent to the stone steps and stand on both sides of the road. The tight structure of them leads to semantic confusion.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussions","content":"\u003cp\u003eThe innovation points in this paper are summarized as follows: (1) multi-source data fusion and dedicated dataset construction. To break the limitations of single data sources, RS, DEM, and OAL images are integrated to identify and interpret the style characteristics of Chinese traditional architectural heritage. According to the list of CTSs, the CTS directory of architectural styles is established, and typical samples are selected and labeled. A set of special datasets suitable for interpretation of style characteristics of traditional architectures is constructed. (2) The perspective of Chinese traditional architectural style. The existing research mostly focuses on the macro settlement pattern or single architectural type, and neglects the landscape features of the architectural styles. With the help of DL technology, this paper quantitatively revealed the representative characteristics and characteristic combination rules for each architectural style, such as environmental characteristics, architectural features and architectural landscape characteristics. (3) Construction of hierarchical analysis framework of \u0026ldquo;architectural environment characteristics - architectural features- architectural landscape characteristics\u0026rdquo;. Natural environment for each architectural style is firstly analyzed, and then architectural features for each style, which include architectural components, roof types, architectural colors and materials, are analyzed by using the classification network and visualization technology. Finally, the representative architectural landscape elements are identified by using semantic segmentation model. This framework elucidates the inherent coupling relationship between natural environment, architecture, and landscape semantics, providing a systematic theoretical basis for the research and protection of traditional architectural heritage.\u003c/p\u003e\u003cp\u003eBased on Grad-CAM visualization analysis, we found that the model not only focuses on the architectural features, but also relies on the surrounding architectural landscape elements, when discriminating architectural styles. Both of them reflect the collaborative shaping of regional environment, cultural influence and functional requirements on traditional architectures. (1) Regional environment. In terms of roof type, the architectures of Min and Chuan have different roof shapes, but both utilize overhanging gable roofs. The southern region, where these architectures are located, has a rainy climate, and the structure of overhanging gable roofs can reduce the erosion of rainwater on the architectures. For the architectural materials, the southern regions are rich in forests, and the architectures of the Su, Hui and Chuan mostly use wood as architectural material. The winter in northern China is cold, dry and windy, and the architectures of Jing and Jin widely use masonry and rammed earth as architectural materials to prevent wind and keep warm. (2) Cultural influence. The architectures of Su construct the natural landscape artistic conception through landscape elements such as water, bridge and stone. The architectures of Hui are deeply influenced by the Confucian culture, and it emphasizes simplicity and solemnity in color. In the architectures of Jing, not only does the shape and decoration of the door strictly comply with etiquette norms, but also stone steps, stone balustrade and gate piers also strengthen the grade through materials and shapes. The architectures of Jin are deeply influenced by farming culture. Vat and stone mill not only meet daily practical needs, but also serve as the material carrier of farming production, which reflect regional economic form and cultural tradition. (3) Functional requirements. In terms of wind and fire prevention, the horsehead walls of the architectures of Hui and its hard mountain roofs highlight the function. For the drainage, both the architectures of Min and Chuan widely use the overhanging gable roofs to achieve effective drainage. For the water security, as the key landscape elements, wells meet the demand of domestic water. It should be noted that some important landscape elements, such as stone carvings, wells, stone mills, are not paid much attention to during the identification. It may be related to insufficient training samples or low element dominance during image acquisition, but they are still the key elements to the identification of architectural style, which carry unique cultural connotation.\u003c/p\u003e\u003cp\u003eThere are still some limitations in our study, which include two aspects: on the one hand, the scale of dataset is smaller, and a small number of images downloaded from the network belong to the wrong architectural styles. It may have a negative impact on the analysis of characteristics of architectural styles. On the other hand, the current model is better at identifying visually significant and structural features, such as roofs and horsehead walls, and not good at the small size of landscape elements that rely on other semantic context, such as censers. To address these limitations, future work can focus on the following aspects: (1) Expanding the high-quality image dataset by incorporating images from professional map platforms, academic databases, and live photos taken by residents. This approach will improve the accuracy and comprehensiveness of the data. (2) Developing more advanced algorithms that integrate geographic location and cultural prior knowledge to enhance the model's understanding of architectural style. (3) Expanding the research scope to include historical images, which can reveal the spatial evolution patterns and internal mechanisms of traditional architectural heritage and settlements.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eBased on RS, DEM and OAL images, style characteristics of Chinese traditional architectures are identified and interpreted using deep learning and Grad-CAM technology. The hierarchical analysis framework of \u0026ldquo;architectural environment characteristics - architectural features- architectural landscape characteristics\u0026rdquo; is constructed, which reveals the architectural characteristics from the perspective of architectural style. Natural environment for each architectural style is firstly analyzed, and then architectural features for each style, which include architectural components, roof types, architectural colors and materials, are analyzed by using the classification network and visualization technology. Finally, the representative architectural landscape elements were identified by using semantic segmentation model. It explains the intrinsic coupling relationship among natural environment, architectures and landscape semantic expression. This research proves that deep learning is an effective method to interpret style characteristics of traditional architectures, offering a novel pathway for the digital conservation and stylistic studies of traditional architectural heritage. In the future, high-quality images, advanced algorithms, and other research on traditional architectures can be further expanded.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003eThis work was supported in part by the Science Foundation of Hebei Normal University under Grant L2025K04, and in part by the Science and Technology Project of Hebei Education Department under Grant BJK2022031.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eC. C.Y., L.S.D., D.J.Y., H.Y.C., and F.Z.Y. collected data, conducted experimental work, and analyzed and verified the results. C.C.Y. and L.S.D. developed the methodology and wrote the manuscript. All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAppendix A is provided within the supplementary information files.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eZeng C, Liu PL, Huang LQ. Measuring the fragmentation of traditional village architectural landscape from the perspective of landscape diversity and heterogeneity. Sci Geogr Sin. 2023;43:1973\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu LY. Zhongguo ren ju shi [History of Chinese Human Settlements]. China Architecture \u0026amp; Building; 2014. pp. 215\u0026ndash;28.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu PL. On Construction and Utilization of Chinese Traditional Settlements Landscape's Genetic Map. Peking University; 2011.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHu Z, Liu PL, Deng YY, Zheng WW. Research on identification and extraction methods of traditional settlement landscape genes. Sci Geogr Sin. 2015;35:1518\u0026ndash;24.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDuan YP, Chen MX, Liu Y, Wang Y, Zhang L. Research on the Cultural Landscape Features and Regional Variations of Traditional Villages and Dwellings in Multicultural Blending Areas: A Case Study of the Jiangxi-Anhui Junction Region. Appl Sci. 2025;15:2185.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRen P, Hong B, Liu Y, Zhou J. Spatial change characteristics of rural settlements and their impacts on landscape patterns based on RS and GIS. Acta Ecol Sin. 2014;34:3331\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang JW. The syntactic graph interpretation of the traditional settlement space from the perspective of cognitive schema: A study of four villages of Han nationality and Zhuang nationality in northern Guangxi. J Beijing Univ Soc Sci. 2024;23:109\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi L. Establishment of characteristic atlas of traditional settlements landscape in multicultural area: Take southern Shaanxi for example. Areal Res Dev. 2020;39:175\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXing Y, Chen HL, Xing YM. A model for landscape heritage information extraction and 3D generation from sketches and text prompts with IoT-enhanced data fusion. Alexandria Eng J. 2025;129:937\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu Y, LSEVGG. An attention mechanism and lightweight-improved VGG network for remote sensing landscape image classification. Alexandria Eng J. 2025;127:943\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMathias M, Martinovic A, Weissenberg J, Haegler S. Automatic architectural style recognition. ISPRS Int Arch Photogramm Remote Sens Spat Inf Sci. 2012;XXXVIII\u0026ndash;5/W16:1\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHan Q, Yin C, Deng YY, Liu PL. Towards classification of architectural styles of Chinese traditional settlements using deep learning: A dataset, a new framework, and its interpretability. Remote Sens. 2022;14:5250.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTan GY, Zhu JK, Chen ZX. Deep learning based identification and interpretability research of traditional village heritage value elements: A case study in Hubei Province. Herit Sci. 2024;12:200.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDai ML, Ward WOC, Meyers G, Tingley DD, Mayfield M. Residential building facade segmentation in the urban environment. Build Environ. 2021;199:107921.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHaznedar B, Bayraktar R, Ozturk AE, Arayici Y. Implementing PointNet for point cloud segmentation in the heritage context. Herit Sci. 2023;11:2.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun RF, Zhang SL, Su MH, Hu LH, Zhang JF. Weakly supervised semantic segmentation for ancient architecture based on multiscale adaptive fusion and spectral clustering. Int J Appl Earth Obs Geoinf. 2025;126:104164.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXiong YZ et al. IEEE,. Accurate detection of historical buildings using aerial photographs and deep transfer learning. In \u003cem\u003eIGARSS 2020\u0026ndash;2020 IEEE International Geoscience and Remote Sensing Symposium\u003c/em\u003e 1592\u0026ndash;1595 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDu LY, Wang YS, Bi. -YOLO: A novel object detection network and dataset for components of China heritage buildings. J Build Eng. 2024;97:110817.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eObeso AM, Vazquez MSG, Acosta AAR, Benois-Pineau J, Connoisseur. Classification of styles of Mexican architectural heritage with deep learning and visual attention prediction. In \u003cem\u003eProceedings of the 15th International Workshop on Content-Based Multimedia Indexing\u003c/em\u003e 1\u0026ndash;7ACM, (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun ML, Zhang F, Duarte F, Ratti C. Understanding architecture age and style through deep learning. Cities. 2022;128:103787.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZu XY, Gao C, Wang Y. Interpreting regional characteristics of Tibetan-Qiang houses in Northwestern Sichuan by deep learning and image landscape. Int J Appl Earth Obs Geoinf. 2024;129:103865.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePan ZK, Xu JS, Guo YB, Hu YM, Wang G. X. Deep learning segmentation and classification for urban village using a worldview satellite image based on U-Net. Remote Sens. 2020;12:1574.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen QT, Zhang L, Duan YP, Fan SQ. Spatial-temporal pattern and evolution of traditional villages in Jiangxi province. Natl Remote Sens Bull. 2021;25:2460\u0026ndash;71.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang G, Liu Z, Van Der Maaten L, Weinberger K. Q. Densely connected convolutional networks. In \u003cem\u003eProceedings of the IEEE Conference on Computer Vision and Pattern Recognition\u003c/em\u003e 2261\u0026ndash;2269IEEE, (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWoo S, Park J, Lee JY, Kweon IS. CBAM: Convolutional block attention module. In Computer Vision \u0026ndash; ECCV 2018. Springer; 2018. pp. 3\u0026ndash;19.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang XX, Li SD, Chen CY. Classification of architectural styles in Chinese traditional settlements using remote sensing images and building facade pictures. J Geogr Sci. 2024;34:2457\u0026ndash;76.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXie EZ et al. Curran Associates,. SegFormer: Simple and efficient design for semantic segmentation with transformers. In \u003cem\u003eAdvances in Neural Information Processing Systems Vol\u003c/em\u003e. 34 12077\u0026ndash;12090 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZheng SX et al. IEEE,. Rethinking semantic segmentation from a sequence-to-sequence perspective with transformers. In \u003cem\u003eProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition\u003c/em\u003e 6877\u0026ndash;6886 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang WH et al. IEEE,. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions. In \u003cem\u003eProceedings of the IEEE/CVF International Conference on Computer Vision\u003c/em\u003e 548\u0026ndash;558 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSelvaraju RR et al. IEEE,. Grad-CAM: Visual explanations from deep networks via gradient-based localization. In \u003cem\u003eProceedings of the IEEE International Conference on Computer Vision\u003c/em\u003e 618\u0026ndash;626 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHinton GE, Roweis S. Stochastic neighbor embedding. Adv Neural Inf Process Syst. 2002;15:833\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu LM. History and Perspective of Development of Rural Landscape Planning in China. J Ecol Rural Environ. 2001;17:52\u0026ndash;5.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePei BJ. The Study about Natural and Geographical Factors on Rural Landscape Spatial Structure in Northeastern Hainan Island. Hainan University; 2015.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHe KM, Zhang XY, Ren SQ, Sun J. IEEE,. Deep residual learning for image recognition. In \u003cem\u003eProceedings of the IEEE Conference on Computer Vision and Pattern Recognition\u003c/em\u003e 770\u0026ndash;778 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSandler M, Howard A, Zhu ML, Zhmoginov A, Chen LJ. Mobilenetv2: Inverted residuals and linear bottlenecks. In \u003cem\u003eProceedings of the IEEE Conference on Computer Vision and Pattern Recognition\u003c/em\u003e 4510\u0026ndash;4520IEEE, (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTan MX, Le QV, Efficientnet. Rethinking model scaling for convolutional neural networks. In \u003cem\u003eProceedings of the 36th International Conference on Machine Learning\u003c/em\u003e 6105\u0026ndash;6114PMLR, (2019).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables 1 and 6","content":"\u003cp\u003eTables 1 and 6 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-heritage-science","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"hsci","sideBox":"Learn more about [Heritage Science](http://heritagesciencejournal.springeropen.com)","snPcode":"40494","submissionUrl":"https://submission.nature.com/new-submission/40494/3","title":"npj Heritage Science","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Architectural style, Chinese traditional architectures, Deep learning, Multi-source data","lastPublishedDoi":"10.21203/rs.3.rs-7971684/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7971684/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFrom the perspective of Chinese traditional architectural style (Su, Hui, Jing, Min, Chuan and Jin), environmental characteristics, architectural features and architectural landscape characteristics of traditional architectures are identified and interpreted using multi-source data, including remote sensing (RS) images, digital elevation model (DEM) data, online architectural landscape (OAL) images. By deep learning (DL) and clustering analysis methods, this paper constructs an hierarchical framework for feature recognition of Chinese traditional architecture, which systematically reveals the inherent coupling relationship between geographical environment, architectural features, and landscape semantics of traditional architectures. The framework can provide technical support for the protection of Chinese traditional architectural cultural heritage.\u003c/p\u003e","manuscriptTitle":"Identifying and Interpreting Traditional Architectural Style Characteristics Based on Deep Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-17 06:35:59","doi":"10.21203/rs.3.rs-7971684/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-11-26T00:21:53+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-25T07:15:06+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-24T07:48:09+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-20T01:15:01+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-17T13:34:46+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"288756821505532893917164892529968650769","date":"2025-11-06T13:44:34+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-06T13:17:47+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"96133740880362088315042100003804402791","date":"2025-11-06T12:56:00+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"71352229730569206199404676798671481519","date":"2025-11-06T07:25:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"178045961006179312737663681255448572354","date":"2025-11-06T05:21:56+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"305044417857500270078641876505406008130","date":"2025-11-06T01:38:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"168222187287040164191068795879255349873","date":"2025-11-06T01:09:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"153612284125498414727021566033979555432","date":"2025-11-06T00:43:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"336243596490864664402991632502648489655","date":"2025-11-06T00:11:50+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-06T00:02:56+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-31T11:07:04+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-31T11:03:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj heritage science","date":"2025-10-28T12:28:04+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-heritage-science","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"hsci","sideBox":"Learn more about [Heritage Science](http://heritagesciencejournal.springeropen.com)","snPcode":"40494","submissionUrl":"https://submission.nature.com/new-submission/40494/3","title":"npj Heritage Science","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"4272f246-f230-474a-8f82-d8266c845ce2","owner":[],"postedDate":"November 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-11T08:23:20+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-17 06:35:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7971684","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7971684","identity":"rs-7971684","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}