Applying Convolutional Neural Networks for Mustard Variety Recognition

Applying Convolutional Neural Networks for Mustard Variety Recognition | 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 Research Article Applying Convolutional Neural Networks for Mustard Variety Recognition LAURA SLEBIODA, BOGNA ZAWIEJA This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4752375/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 16 Jan, 2025 Read the published version in Euphytica → Version 1 posted 13 You are reading this latest preprint version Abstract The aim of this study was to develop and apply a Convolutional Neural Network (CNN) model to recognize and classify white mustard ( Sinapis alba L. ) varieties, addressing the complex task of discriminating among 57 varieties. Utilizing a one-dimensional CNN model, the research focused on multivariate analysis based on a set of 15 traits. The CNN architecture included convolutional layers, batch normalization, pooling, flattening, dropout, and dense layers. The model demonstrated effectiveness in classifying varieties, achieving high accuracy and providing valuable insights into potential new varieties. Subset division, a new approach, was applied. Evaluation metrics, including accuracy, F1 score, precision, and recall, were calculated for eight subsets, confirming the model's robust performance. Comparison with existing studies highlights the uniqueness of the proposed algorithm in addressing variety discrimination. The approach contributes to agricultural advancements, offering a reliable tool for breeders to assess variety distinctness and streamline the testing process. The model's ability to detect unknown varieties further enhances its utility in agricultural research covering a comprehensive and impactful advancement in variety classification. Agricultural Algorithms Classification Convolutional Neural Networks Mustard Variety Recognition Figures Figure 1 Figure 2 Figure 3 1. Introduction The continually growing demand for plant products entails the necessity of breeding new plant varieties that are better adapted to environmental conditions, resistant to diseases, and yielding high crops. Therefore, breeders continually develop new varieties that, after years of research, can be registered. In Poland, breeders submit new varieties to the Research Centre for Cultivar Testing (COBORU). The basis for registration is demonstrating the Distinctness, Uniformity, and Stability (DUS) of the new variety. Variety evaluations are based on two or three years of experiments, where new varieties are sown alongside all known and registered varieties (whose similarity has been indicated by the breeder). Internationally, the coordination of DUS research is managed by the International Union for the Protection of New Varieties of Plant (Union Internationale pur la Protection des Obtentions Vegetates – UPOV), headquartered in Geneva. Distinctness, uniformity, and stability assessments of new varieties traditionally require demonstrating distinctness in at least one attribute. As mentioned above, the new variety must be distinct from known, already registered varieties and other new varieties sown in the same experiment. One-dimensional analysis of variance is conducted separately for each attribute. These principles are recommended by UPOV. However, before conducting separate analyses, it would be worthwhile to apply multivariate methods to determine whether a given variety differs from others simultaneously in all examined attributes. Multivariate evaluations are crucial for selections as they allow for a comprehensive analysis of multiple traits simultaneously, thereby providing a more holistic assessment of a variety's performance (Sharma, 2006 ). If such an analysis shows distinctness of the new varieties, one-dimensional analyses can then be performed to identify specific distinguishing features. If, however, multivariate analysis does not show distinctness of any of the new varieties, there is no need to conduct one-dimensional analyses. Multivariate analysis could be useful for evaluating varieties submitted for registration and by breeders allowing for the evaluation of the novelty of the obtained variety and its chances of a positive outcome in DUS testing. This could eliminate varieties that do not show distinctness from further testing. Yield could also be added to the set of analyzed traits, allowing breeders to identify promising varieties. Mustard is an annual plant belonging to the cabbage family, cultivated for various purposes. These species are often sown as an aftercrop, providing green manure (Alferov and Chernova, 2018 ). Young green plants are valuable fodder for livestock. The seeds are toxic to animals but serve as a culinary spice for humans and for mustard production (Grygier, 2023 ). However, large quantities of seeds are also harmful to humans. White mustard ( Sinapis alba L. ) is used in medicine, such as in the treatment of stomach inflammations, acidity, ulcers, constipation, respiratory diseases (Guo et al., 2013 ) and many others (Grygier, 2023 ). In this study, a convolutional neural network (CNN) was used to demonstrate the distinctness of new white mustard varieties. This is a multivariate analysis based on a model that leverages mathematical linear operations between matrices (Albawi, et al., 2017 ). A CNN model mimics the neurons of the human nervous system. Neural network models take multidimensional data (input data) and through hidden layers, they can output the classification of analyzed objects. CNNs use matrix multiplication in the hidden layers to detect patterns in images or data. The core components of a CNN include convolutional layers, pooling layers, and fully connected layers (Li et al., 2021 ). Convolutional layers apply filters to the input data to create feature maps, which highlight important features of the data such as edges, textures, and shapes. Pooling layers reduce the dimensionality of these feature maps, retaining the most critical information while reducing computational load and mitigating the risk of overfitting. Fully connected layers, positioned towards the end of the network, take the high-level filtered data and perform the final classification (Zhang et al., 2018 ). The network's performance depends on the number of its levels. However, as the number of levels increases, the time required for training and testing the network also increases. Therefore, the number of levels should be adjusted to the specific case. 2D-CNN networks, as a biologically inspired model of deep learning, were initially used for image classification, face identification, and pattern recognition (Azizjon, et al., 2020 ). Convolution operations automatically extract complex feature patterns from the image using a filter (kernel). In this study, this method was applied to one-dimensional data, meaning 1D-CNN networks were used, consistent with the dimension of white mustard data. Due to the low values of the evaluation metrics, subset division was applied. In this approach, the original dataset is divided into several subsets. This idea was developed to address challenges associated with distinguishing closely related varieties and improving model sensitivity to specific distinctions. It is a novel concept aimed at helping in situations where the data is limited, making it easier for models to learn from smaller, more focused subsets. While dividing data into subsets is a widely recognized technique in machine learning classification, its application in this study is unique. Abuassba et al. ( 2017 ) demonstrated the benefits of subset division by splitting the dataset and applying various classification methods to each subset, ultimately identifying the most effective method for each. However, the approach employed in this study goes beyond their methodology, applying the subset division in a novel context to enhance classification accuracy. The aim of this work is to develop and apply a CNN model to recognize mustard varieties and to distinguish new mustard varieties from existing ones. Applying a novel subset division approach is to address challenges associated with distinguishing closely related plant varieties, thereby improving the model's sensitivity to specific distinctions. The developed algorithm can be widely used in agriculture, particularly in optimizing the registration and cultivation of mustard varieties. 2. Materials and methods 2.1. Experimental Data The experimental data used in the multivariate analysis are derived from a DUS experiment with white mustard conducted at COBORU in 2009, involving the examination of 57 varieties. The experiments were carried out in a complete randomized block design with two blocks, and for each variety, a sample of a total of 60 plants was taken. 30 plants were randomly selected from each plot. In this experiment, measurements were taken for 15 quantifiable traits (13 continuous and 2 categorical). The list of analyzed traits is provided in Table 1 below. Table 1 List of analyzed traits. Trait Name Seedling: Length of cotyledons (cm) Seedling: Width of cotyledons (cm) Plant: Overall length of stem (cm) Leaf - Number of leaflets Leaf: Length of leaf (cm) Leaf: Width of leaf (cm) Leaf: Length of leaf petiole (cm) Plant: Height at flowering (cm) Flower: Length of flower petal (cm) Flower: Width of flower petal (cm) Siliqua: Length of siliqua (excluding beak) (mm) Siliqua: Length of beak (mm) Siliqua: Width of siliqua (mm) Siliqua: Length of pedicel (mm) Siliqua - Number of seeds in siliqua 2.2. Model description In this study, a Conv1D model, i.e., a one-dimensional convolutional neural network, was employed. Conv1D creates a convolution kernel that is convolved with the input data in one spatial dimension to generate results (Dewantara et al., 2020 ). Batch Normalization was used in a model for classification purposes. This layer introduces random perturbation to the decision boundary of deep networks, forcing the model to learn boundaries with increased margins to the nearest training samples (Balestriero and Baraniuk, 2022 ). MaxPooling1D was used to perform pooling operations with the maximum values in a single spatial dimension when creating output data (Dewantara et al., 2020 ). The flattening layer was utilized as it converts an array into a vector. The Dropout layer was applied as a method to overcome overfitting, which often occurs when using a deep learning approach. The Dropout layer randomly reduces the number of neuron units in the network to decrease the number of connections in each iteration of the learning process. The dense layer that was used, is a simple layer of neurons, where each neuron receives input data from all neurons in the previous layer (Dumane, 2020 ). The model utilized in this study is a sequential model. The sequential model applied to the analyzed data for mustard varieties is presented in Table 2 . It was initialized using the Keras library (an open-source Python library for machine learning). The layers of the model include two convolutional layers (Conv1D), two Batch Normalization layers, a pooling layer (MaxPooling1D), a flattening layer (Flatten), a Dropout layer, and two Dense layers. Each layer has its own parameters such as size, activation function, and input shape. In total, the model has 2,408 parameters, of which 2,344 are trainable, and 64 are not, as they are parameters of batch normalization layers. The activation function in layers 1, 3, and 7 is a rectified linear unit (ReLU), shown in Fig. 1 . If the input value is below zero, the output value is zero. When the input value rises above a certain threshold, it has a linear relationship with the dependent variable (Bisen, 2021 ). The formula for the ReLU function is provided below (Formula 1). The activation function in layer 9 is Softmax. This function is a combination of multiple sigmoid functions (Sharma, et al., 2020 ). Since we know that the sigmoid function returns values in the range from 0 to 1, they can be treated as probabilities of data points for a specific class. Unlike sigmoid functions used for binary classification, the Softmax function can be used for multi-class classification problems. The function returns the probability for each data point for all individual classes. When building a network or model for multi-class classification, the output layer of the network will have the same number of neurons as the number of classes in the target object. The formula for the Softmax function is provided below (Formula 2). Table 2 Model summary (Model: "sequential"). Layer Number Layer Output Shape Parameter 1 conv1d (None, 13, 16) 64 2 batch_normalization (None, 13, 16) 64 3 conv1d (None, 11, 16) 784 4 batch_normalization (None, 11, 16) 64 5 max_pooling1d (None, 5, 16) 0 6 flatten (None, 80) 0 7 dense (None, 16) 1296 8 dropout (None, 16) 0 9 dense_1 (None, 8) 136 Total parameters : 2,408 Trainable parameters : 2,344 Non-trainable params : 64 Formula 1 . ReLU activation function [Bisen, 2021 ]. $$\:f\left(x\right)=\left\{\begin{array}{c}0,\:\:x<0\\\:x,\:\:x\ge\:0\end{array}\right.$$ Where x is the input to the ReLU activation function. Formula 2 . Softmax activation function [Sharma, et al., 2020 ]. $$\:\sigma\:{\left(z\right)}_{j}=\frac{{e}^{{z}_{j}}}{{\sum\:}_{k=1}^{K}{e}^{{z}_{k}}}\:\:\:\:for\:j=1,\dots\:,K.$$ Where: \(\:\sigma\:{\left(z\right)}_{j}\) is the probability for every data point of all the individual classes, \(\:{\left(z\right)}_{j}\) is the input to the Softmax activation function corresponding to the j -th element. In the model, applied loss function was 'categorical_crossentropy,' the optimizer was 'adam,' and the metric was 'accuracy.' The training parameters of the model were configured as follows: the 'epochs' parameter, determining the number of training epochs, is set to 80; the 'batch_size' parameter, defining the size of the training data batches, is set to 50; and applied 'validation_data' technique uses test data as the validation set during training. The results for the applied algorithm were not satisfactory, so cross-validation was applied but the results did not improve. It was then decided to use a new approach – division into subsets. It turned out that the best solution is to divide it into 8 subsets as shown in Fig. 2 . The data was divided into eight subsets as presented in Table 3 . In the subset number 8 were eight varieties and seven varieties in each of the remaining subsets. The aim was to improve the classification accuracy of a complex dataset by creating smaller, more manageable subsets. Table 3 Division of the variety dataset into subsets. Subset Number Varieties 1 1, 2, 3, 4, 5, 6, 7 2 8, 9, 11, 13, 14, 15, 16 3 17, 18, 20, 21, 8000, 8001, 8023 4 8002, 8003, 8006, 8008, 8009, 8010, 8012 5 8007, 8011, 8013, 8014, 8016, 8017, 8018 6 19, 8019, 8020, 8021, 8022, 8024, 8025 7 8026, 8027, 8028, 8029, 8030, 8031, 8032 8 8033, 8034, 8035, 8036, 8037, 8038, 8039, 8040 The detailed operation of data subdivision for the data used in this study is presented in Fig. 3 . First, the user inputs the total number of subsets to be processed (variable n ) which in our case is eight. Then, the loop counter i is initialized to 1, indicating the start of the loop for processing each subset. For the current subset, a CNN model is applied. This step involves training the model using the subset data. Next, two types of plots are generated and shown. First plot is Train and Validation Accuracy which shows the accuracy of the model on both the training and validation data over epochs. Second plot is Confusion Matrix for the predictions, showing how well the model performed in classifying the data. Then, the evaluation metrics (e.g., accuracy, F1 score, precision, recall) for the current subset are printed. The classification decision for new data is printed, indicating whether the new data is classified as an existing variety or a new variety. A decision point checks if the loop counter i equals n , indicating whether all subsets have been processed. If not, the loop counter i is incremented by 1 and the process loops back to apply the CNN model to the next subset. If i is equal to n , the process proceeds to the final step and the mean evaluation metrics for all subsets are printed and also the final classification decision based on the collective results of all subsets is printed. The computations were conducted in Python 3.7.7, utilizing libraries such as tensorflow, scikit-learn, pandas, numpy, scipy, matplotlib, and seaborn. 2.3. Application of the model to the data Subsequently, the created Convolutional Neural Network (CNN) model in the Keras library was applied to each subset. The input data consists of 15 features describing a mustard variety, and the output label is an assigned variety number. Using the StandardScaler method from the sklearn.preprocessing library, the input data was standardized, ensuring that mean is equal to 0 and variance is equal to 1. This standardization process is achieved using the fit_transform() method, which calculates mean and standard deviation for each feature based on the available training data and then transforms the input data by subtracting the mean and dividing by the standard deviation. This process was applied to enhance the stability and effectiveness of machine learning. Subsequently, using the ‘to_categorical’ function from the tensorflow.keras.utils library, the output labels are transformed into binary form through a process known as one-hot encoding. This transformation ensures that each label is represented as a binary vector with a length equal to the number of classes, which is require for training a neural network. Thereafter, the data is divided into a training set and a test set using the ‘train_test_split’ function from the sklearn.model_selection library. The parameter ‘test_size = 0.2’ indicates that 20% of the data will constitute the test set. The parameter ‘stratify = y’ ensures that the split maintains the class proportions in the output labels. The parameter ‘random_state = 42’ sets the seed for randomness to ensure reproducibility of the split. The model is evaluated separately for each subset and in the end, mean is calculated for all evaluation metrics. Additionally, accuracy is calculated on the test set. The evaluation metrics include: train accuracy, test accuracy, F1 score, precision and recall. The model also assumes the possibility of the existence of previously unknown varieties. When providing the results of assigning the given data to a specific variety, the model may recognize these data as a new variety. Therefore, when submitting data to determine the variety, at least 10 observations should be provided. The model will iterate through all subsets, assign each observation to a specific variety, and then check if each observation is assigned to the same variety. If not, it means that the data is classified as a new variety. After checking this for all subsets, if in any of them, all given observations are classified into one specific variety, then the model classifies the given data as that variety. If in every subset the result is 'The new data is classified as a new variety.', it means that the data belongs to a new, previously unknown variety. 3. Results Using the trained model, predictions are made on the new data. An experiment was conducted with data related to variety 8012 (10 observations randomly selected from 60 observations). These data were not used previously as training data. The results for each subset are presented in Table 4 . Train accuracy ranges from 0.905 to 0.975 and test accuracy ranges from 0.731 to 0.900. The subset where classification for variety 8012 was examined is subset 4. The decision in this subset is: 'The new data belongs to variety: 8012'. Table 4 Results for input data of variety 8012. Subset number Train Accuracy Test Accuracy F1 Score Precision Recall Decision 1 0.956 0.731 0.734 0.742 0.731 The new data is classified as a new variety. 2 0.975 0.843 0.840 0.845 0.843 The new data is classified as a new variety. 3 0.906 0.789 0.793 0.806 0.789 The new data is classified as a new variety. 4 0.970 0.900 0.899 0.905 0.900 The new data belongs to variety: 8012 5 0.943 0.774 0.775 0.788 0.774 The new data is classified as a new variety. 6 0.905 0.881 0.876 0.877 0.881 The new data is classified as a new variety. 7 0.945 0.854 0.851 0.957 0.854 The new data is classified as a new variety. 8 0.933 0.824 0.825 0.847 0.824 The new data is classified as a new variety. For each subset, charts comparing train accuracy with validation accuracy were made. These charts are presented in Table 5 . The model's accuracy varies depending on the number of epochs and regularization that was conducted using dropout layer. The charts show that there are no noteworthy differences between the compared accuracies, indicating that overfitting did not occur. Additionally, a confusion matrix was created for each subset, as shown in Table 6 . It can be observed that the correct prediction ranges from 58–100% (82% on average), with the highest accuracy in subset 4 ranging from 78–100% (90% on average). Mean values for the evaluation metrics are presented in Table 7 . Ultimately, the model classified the data as belonging to variety 8012. This indicates that the model is functioning correctly. Table 7 Mean values for the evaluation metrics and decision. Parameter Mean Mean Train Accuracy 0.942 Mean Accuracy 0.824 Mean F1 Score 0.824 Mean Precision 0.833 Mean Recall 0.824 Decision The new data belongs to variety: 8012 4. Discussion Lee et al. ( 2015 ) conducted a study on convolutional neural networks (CNN) to learn unsupervised feature representations for 44 different plant species. They concluded that learning features through CNN can provide superior feature representation for leaf images compared to hand-crafted features. Saini, et al. ( 2020 ) presented a method for classifying healthy and diseased apple and grape leaves using a convolutional neural network. The network achieved a training accuracy of 99.96% and a testing accuracy of 99.90%. Kattenborn, et al. ( 2019 ) applied a segmentation approach based on CNN for detailed mapping of plant species communities. The demonstrated approach achieved segmentation and mapping accuracy of at least 84%. Wang, et al. ( 2021 ) constructed a multiscale CNN model with attention (AMSCNN) to enhance multiscale feature extraction capabilities. Proposed model was applied to plant species identification and achieved an accuracy of 73.62%. Ashturkar and Bhalchandra ( 2021 ) proposed a designed CNN model and AlexNet model with application-oriented fine-tuning to classify plant species based on leaves. They achieved 91.19% accuracy for their CNN model and 97.41% accuracy for the AlexNet model. Wang and Su ( 2022 ) compared CNN models used for detecting the phenotype of four cereal crops. They selected the latest CNNs related to organ counting, weed segmentation, biotic and abiotic stress assessment, and seed variety classification. Results showed the importance of neural network architecture, development strategy, and annotated datasets in designing models for different tasks, directly impacting CNN performance. Despite limited phenotypic information on seeds, the proposed CNN models were able to effectively classify seeds with an accuracy exceeding 95%. The authors also observed that deep CNN models have higher accuracy than shallow networks. Zhu et al. ( 2018 ) utilized deep CNN for plant species classification, testing existing classifiers and creating a new one for leaf, flower, fruit, stem, and branch classification. The lowest accuracy was obtained for stem classification, where the new classifier achieved the best result at 65.2% and existing classifiers ranged from 44.8–58.7%. The highest accuracy was obtained for fruit classification, with accuracy ranging from 44.8–66.3% for known CNN algorithms and 90.2% for the newly created model. Alom et al. ( 2023 ) employed CNN for the classification of two types of B. Napus species, B. Rapa and B. Alba, considering flower, leaf, and packet features individually. The DenseNet201 model achieved the best results, getting 100% accuracy for flowers, 97% for leaves, and 100% for packets. The referenced studies focused on distinguishing between plant species, while this study specifically addresses the discrimination of plant varieties, which is more complex. The presence of a large number of varieties makes classification more challenging and in this study the number of varieties is 57. Morphological characteristics are unstable, they are created as a result of an environmental and climatic conditions, making phenotypic plasticity a consequence of adaptation (Korir et al., 2013 ). Therefore, molecular markers, statistical tests, and software are employed for classification and characterization of plant varieties. As noted by Korir et al. ( 2013 ), certain limitations create challenges to the development and application of modern scientific methods in identifying varieties of many cultivated plants. Taking all of this into consideration, the algorithm developed in this study classifies white mustard varieties very effectively and moreover, informs if a given data belongs to a previously unknown, new variety. Dividing data into subsets turned out to be an efficient method. Distinguishing 57 varieties at once is exceptionally difficult, so the implementation of a new idea, dividing the data into subsets, made this task easier. In studies on distinguishing varieties, the number of varieties in the dataset is usually much smaller than 57. Koklu, et al. ( 2021 ) studied five rice varieties and achieved 100% classification success with CNN. Wang and Song ( 2023 ) examined five varieties of sweet maize seeds using neural networks and achieved 95% accuracy. While other studies typically dealt with datasets containing fewer varieties, this approach tackles a larger and more complex dataset. The developed algorithm can be widely used in agriculture, at all stages of cultivation. Thanks to this solution, varieties submitted for registration have the opportunity to achieve positive results in DUS tests and non-promising varieties could be eliminated from further research in terms of distinctness. In future studies, the proposed approach can be expanded by adding new varieties to the dataset and adjusting the parameters of the CNN architecture for these new varieties. Declarations Author Contribution L.S. did conceptualization, methodology, developed software and prepared all figures. B.Z. was responsible for data curation and supervision. L.S. wrote original draft and B.Z. reviewed and edited it.All authors approved the version to be published. Data Availability Data will be available on request. References Abuassba AOM, Zhang D, Luo X, Shaheryar A, Ali H (2017) Improving Classification Performance through an Advanced Ensemble Based Heterogeneous Extreme Learning Machines. Comput Intell Neurosci. 10.1155/2017/3405463. Albawi S, Mohammed TA, Al-Zawi S (2017) Understanding of a convolutional neural network. 2017 International Conference on Engineering and Technology (ICET). 10.1109/icengtechnol.2017.8308186 Alferov A A, Chernova LS (2018) Influence of Fertilizers, Biomass of White Mustard, and Biopreparation Rhizoagrin on Yield and Grain Quality of Spring Wheat ( Triticum aestivum L. ) and Sustainability of the Agroecosystem. Russian Agriculture Sciences 44: 546–550. https://doi.org/10.3103/S1068367418060022 Alom M, Ali MY, Islam MT, Uddin AH, Rahman W (2023) Species classification of brassica napus based on flowers, leaves, and packets using deep neural networks. Journal of Agriculture and Food Research 14. https://doi.org/10.1016/j.jafr.2023.100658 Ashturkar HH, Bhalchandra AS (2021) Plant Species Identification Using Convolutional Neural Network. Intelligent Data Communication Technologies and Internet of Things. Lecture Notes on Data Engineering and Communications Technologies. 10.1007/978-981-15-9509-7 Azizjon M, Jumabek A, Kim W (2020) 1D CNN based network intrusion detection with normalization on imbalanced data. 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC). 10.1109/icaiic48513.2020.9064976 Balestriero R, Baraniuk RG (2022) Batch normalization explained. arXiv. https://doi.org/10.48550/arXiv.2209.14778 Bisen D (2021) Deep convolutional neural network based plant species recognition through features of leaf. Multimedia Tools and Applications 80:6443–6456. https://doi.org/10.1007/s11042-020-10038-w Dewantara DS, Budi I, Ibrohim MO (2020) 3218IR at SemEval-2020 Task 11: Conv1D and Word Embedding in Propaganda Span Identification at News Articles. Proceedings of the 14th International Workshop on Semantic Evaluation 1716–1721. 10.18653/v1/2020.semeval-1.225 Dumane G (2020) Introduction to Convolutional Neural Network (CNN) using Tensorflow. Towards Data Science. https://towardsdatascience.com/introduction-to-convolutional-neural-network-cnn-de73f69c5b83 [available: 13.11.2023] Grygier A (2023) Mustard Seeds as a Bioactive Component of Food. Food Reviews International 39(7):4088-4101. 10.1080/87559129.2021.2015774 Guo X, Lu H, Lin Y, Chen B, Wu C, Cui Z, Wang Y, Xu Y (2013) Skin penetration of topically applied white mustard extract and its effects on epidermal Langerhans cells and cytokines. International Journal of Pharmaceutics 457(1):136-142. 10.1016/j.ijpharm.2013.09.015. Kattenborn T, Eichel J, Fassnacht FE (2019) Convolutional Neural Networks enable efficient, accurate and fine-grained segmentation of plant species and communities from high-resolution UAV imagery. Scientific Reports 9(17656). https://doi.org/10.1038/s41598-019-53797-9 Koklu M, Cinar I, Taspinar YS (2021) Classification of rice varieties with deep learning methods. Computers and Electronics in Agriculture 187. https://doi.org/10.1016/j.compag.2021.106285. Korir NK, Han J, Shangguan L, Wang C, Kayesh E, Zhang Y, Fang J (2013) Plant variety and cultivar identification: advances and prospects. Critical Reviews in Biotechnology 33(2):111–125. 10.3109/07388551.2012.675314 Lee SH, Chan CS, Wilkin P, Remagnino P (2015) Deep-plant: Plant identification with convolutional neural networks. 2015 IEEE International Conference on Image Processing (ICIP). 10.1109/icip.2015.7350839 Li Z, Liu F, Yang W, Peng S, Zhou J (2021) A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. IEEE Transactions on Neural Networks and Learning Systems: 1–21. 10.1109/tnnls.2021.3084827 Saini G, Khamparia A, Luhach AK (2020) Classification of Plants Using Convolutional Neural Network. First International Conference on Sustainable Technologies for Computational Intelligence. Advances in Intelligent Systems and Computing. 10.1007/978-981-15-0029-9 Sharma JR (2006) Statistical and Biometrical Techniques in Plant Breeding. New Age International, New Delhi. Sharma S, Sharma S, Athaiya A (2020) Activation functions in neural networks. International Journal of Engineering Applied Sciences and Technology 4(12):310–316. Wang X, Zhang C, Zhang S (2021) Multiscale Convolutional Neural Networks with Attention for Plant Species Recognition. Computational Intelligence and Neuroscience. 10.1155/2021/5529905 Wang Y, Song S (2023) Variety identification of sweet maize seeds based on hyperspectral imaging combined with deep learning. Infrared Physics & Technology 130. https://doi.org/10.1016/j.infrared.2023.104611. Wang Y-H, Su W-H (2022) Convolutional Neural Networks in Computer Vision for Grain Crop Phenotyping: A Review. Agronomy 12(11). https://doi.org/10.3390/agronomy12112659 Zhang Z, Zhao Y, Liao X, Shi W, Li K, Zou Q, Peng S (2018) Deep learning in omics: a survey and guideline. Briefings in Functional Genomics. 10.1093/bfgp/ely030 Zhu H, Liu Q, Qi Y, Huang X, Jiang F, Zhang S (2018) Plant identification based on very deep convolutional neural networks. Multimedia Tools and Applications 7. 10.1007/s11042-017-5578-9 Tables Tables 5 and 6 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Tables5and6.docx Cite Share Download PDF Status: Published Journal Publication published 16 Jan, 2025 Read the published version in Euphytica → Version 1 posted Editorial decision: Revision requested 20 Dec, 2024 Reviews received at journal 12 Dec, 2024 Reviews received at journal 12 Dec, 2024 Reviews received at journal 11 Dec, 2024 Reviews received at journal 11 Dec, 2024 Reviewers agreed at journal 29 Nov, 2024 Reviewers agreed at journal 28 Nov, 2024 Reviewers agreed at journal 28 Nov, 2024 Reviewers agreed at journal 28 Nov, 2024 Reviewers invited by journal 20 Jul, 2024 Editor assigned by journal 16 Jul, 2024 Submission checks completed at journal 16 Jul, 2024 First submitted to journal 16 Jul, 2024 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-4752375","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":331696737,"identity":"484092a5-bd88-4d93-aea9-93373a8182f2","order_by":0,"name":"LAURA SLEBIODA","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFklEQVRIiWNgGAWjYBACPigtwyDB2MDAw2CDLMncgE0LG5TmgWpJQ5ZkJKQFTB4mQgv7GbMPPxjqePilmxsfvKk5n7idgffgw5877PIYpBuxa+HJMZ7Zw3CYR3LOwWbDOcduJ+5s4Es2kDyTXMwgcxCHw3KMge45wGNwI7FNmoftduKGAzxmEoZtzIkNEonYtfC/MWb8A3SYPVjLv3MQLYlt9bi1SOQYM/MwMPMYAJVJ87YdgGg52HYYj5ZnxcwyBod5JO4A/TK3L9l4ZzNfsmFj2/HENhx+4edP3sz4pqJOjn92+8MHb77ZyW5n7wWGWFt1Yr908wFsWiDAAJkNdCfUAbg1oGuHamEgWssoGAWjYBQMcwAAeh5bBrzF3zIAAAAASUVORK5CYII=","orcid":"","institution":"Poznań University of Live Sciences","correspondingAuthor":true,"prefix":"","firstName":"LAURA","middleName":"","lastName":"SLEBIODA","suffix":""},{"id":331696738,"identity":"57f16ba0-fdc9-4a9c-986d-0632836c2cb4","order_by":1,"name":"BOGNA ZAWIEJA","email":"","orcid":"","institution":"Poznań University of Live Sciences","correspondingAuthor":false,"prefix":"","firstName":"BOGNA","middleName":"","lastName":"ZAWIEJA","suffix":""}],"badges":[],"createdAt":"2024-07-16 21:23:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4752375/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4752375/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10681-025-03461-3","type":"published","date":"2025-01-16T15:57:19+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":62094668,"identity":"d42b05a2-63f1-4fe5-9666-c29e99898a77","added_by":"auto","created_at":"2024-08-09 08:23:11","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":12910,"visible":true,"origin":"","legend":"\u003cp\u003eReLU activation function [Bisen, 2021].\u003c/p\u003e","description":"","filename":"figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4752375/v1/e95bc3cbdfbbb3595dc6fbeb.png"},{"id":62095352,"identity":"a6ee597a-2276-4479-8460-1e8cc28628b8","added_by":"auto","created_at":"2024-08-09 08:31:11","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":31138,"visible":true,"origin":"","legend":"\u003cp\u003eThe division of Dataset into Subsets.\u003c/p\u003e","description":"","filename":"figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4752375/v1/4bc6539a3050249cb8460fd7.png"},{"id":62094670,"identity":"f6671e2c-f95a-4b69-ab1c-719eaf3aae59","added_by":"auto","created_at":"2024-08-09 08:23:11","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":60795,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of data subdivision.\u003c/p\u003e","description":"","filename":"figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4752375/v1/a880d26d85e0eda281ea1266.png"},{"id":74285574,"identity":"0fca1734-c35c-40ef-a6f7-80281e2bcf6a","added_by":"auto","created_at":"2025-01-20 16:14:11","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":682304,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4752375/v1/c877dffd-7149-4144-add9-a62536de03b3.pdf"},{"id":62094667,"identity":"07abaf8d-6eaf-4682-9ef6-3aa927d93417","added_by":"auto","created_at":"2024-08-09 08:23:11","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":768898,"visible":true,"origin":"","legend":"","description":"","filename":"Tables5and6.docx","url":"https://assets-eu.researchsquare.com/files/rs-4752375/v1/816e422cfe1b7a521b9ce9bc.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Applying Convolutional Neural Networks for Mustard Variety Recognition","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe continually growing demand for plant products entails the necessity of breeding new plant varieties that are better adapted to environmental conditions, resistant to diseases, and yielding high crops. Therefore, breeders continually develop new varieties that, after years of research, can be registered. In Poland, breeders submit new varieties to the Research Centre for Cultivar Testing (COBORU). The basis for registration is demonstrating the Distinctness, Uniformity, and Stability (DUS) of the new variety. Variety evaluations are based on two or three years of experiments, where new varieties are sown alongside all known and registered varieties (whose similarity has been indicated by the breeder). Internationally, the coordination of DUS research is managed by the International Union for the Protection of New Varieties of Plant (Union Internationale pur la Protection des Obtentions Vegetates \u0026ndash; UPOV), headquartered in Geneva.\u003c/p\u003e \u003cp\u003eDistinctness, uniformity, and stability assessments of new varieties traditionally require demonstrating distinctness in at least one attribute. As mentioned above, the new variety must be distinct from known, already registered varieties and other new varieties sown in the same experiment. One-dimensional analysis of variance is conducted separately for each attribute. These principles are recommended by UPOV. However, before conducting separate analyses, it would be worthwhile to apply multivariate methods to determine whether a given variety differs from others simultaneously in all examined attributes. Multivariate evaluations are crucial for selections as they allow for a comprehensive analysis of multiple traits simultaneously, thereby providing a more holistic assessment of a variety's performance (Sharma, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). If such an analysis shows distinctness of the new varieties, one-dimensional analyses can then be performed to identify specific distinguishing features. If, however, multivariate analysis does not show distinctness of any of the new varieties, there is no need to conduct one-dimensional analyses. Multivariate analysis could be useful for evaluating varieties submitted for registration and by breeders allowing for the evaluation of the novelty of the obtained variety and its chances of a positive outcome in DUS testing. This could eliminate varieties that do not show distinctness from further testing. Yield could also be added to the set of analyzed traits, allowing breeders to identify promising varieties.\u003c/p\u003e \u003cp\u003eMustard is an annual plant belonging to the cabbage family, cultivated for various purposes. These species are often sown as an aftercrop, providing green manure (Alferov and Chernova, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Young green plants are valuable fodder for livestock. The seeds are toxic to animals but serve as a culinary spice for humans and for mustard production (Grygier, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, large quantities of seeds are also harmful to humans. White mustard (\u003cem\u003eSinapis alba L.\u003c/em\u003e) is used in medicine, such as in the treatment of stomach inflammations, acidity, ulcers, constipation, respiratory diseases (Guo et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) and many others (Grygier, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn this study, a convolutional neural network (CNN) was used to demonstrate the distinctness of new white mustard varieties. This is a multivariate analysis based on a model that leverages mathematical linear operations between matrices (Albawi, et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). A CNN model mimics the neurons of the human nervous system. Neural network models take multidimensional data (input data) and through hidden layers, they can output the classification of analyzed objects. CNNs use matrix multiplication in the hidden layers to detect patterns in images or data. The core components of a CNN include convolutional layers, pooling layers, and fully connected layers (Li et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Convolutional layers apply filters to the input data to create feature maps, which highlight important features of the data such as edges, textures, and shapes. Pooling layers reduce the dimensionality of these feature maps, retaining the most critical information while reducing computational load and mitigating the risk of overfitting. Fully connected layers, positioned towards the end of the network, take the high-level filtered data and perform the final classification (Zhang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The network's performance depends on the number of its levels. However, as the number of levels increases, the time required for training and testing the network also increases. Therefore, the number of levels should be adjusted to the specific case. 2D-CNN networks, as a biologically inspired model of deep learning, were initially used for image classification, face identification, and pattern recognition (Azizjon, et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Convolution operations automatically extract complex feature patterns from the image using a filter (kernel). In this study, this method was applied to one-dimensional data, meaning 1D-CNN networks were used, consistent with the dimension of white mustard data.\u003c/p\u003e \u003cp\u003eDue to the low values of the evaluation metrics, subset division was applied. In this approach, the original dataset is divided into several subsets. This idea was developed to address challenges associated with distinguishing closely related varieties and improving model sensitivity to specific distinctions. It is a novel concept aimed at helping in situations where the data is limited, making it easier for models to learn from smaller, more focused subsets. While dividing data into subsets is a widely recognized technique in machine learning classification, its application in this study is unique. Abuassba et al. (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) demonstrated the benefits of subset division by splitting the dataset and applying various classification methods to each subset, ultimately identifying the most effective method for each. However, the approach employed in this study goes beyond their methodology, applying the subset division in a novel context to enhance classification accuracy.\u003c/p\u003e \u003cp\u003eThe aim of this work is to develop and apply a CNN model to recognize mustard varieties and to distinguish new mustard varieties from existing ones. Applying a novel subset division approach is to address challenges associated with distinguishing closely related plant varieties, thereby improving the model's sensitivity to specific distinctions. The developed algorithm can be widely used in agriculture, particularly in optimizing the registration and cultivation of mustard varieties.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Experimental Data\u003c/h2\u003e \u003cp\u003eThe experimental data used in the multivariate analysis are derived from a DUS experiment with white mustard conducted at COBORU in 2009, involving the examination of 57 varieties. The experiments were carried out in a complete randomized block design with two blocks, and for each variety, a sample of a total of 60 plants was taken. 30 plants were randomly selected from each plot. In this experiment, measurements were taken for 15 quantifiable traits (13 continuous and 2 categorical). The list of analyzed traits is provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e below.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eList of analyzed traits.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"1\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrait Name\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSeedling: Length of cotyledons (cm)\u003c/p\u003e \u003cp\u003eSeedling: Width of cotyledons (cm)\u003c/p\u003e \u003cp\u003ePlant: Overall length of stem (cm)\u003c/p\u003e \u003cp\u003eLeaf - Number of leaflets\u003c/p\u003e \u003cp\u003eLeaf: Length of leaf (cm)\u003c/p\u003e \u003cp\u003eLeaf: Width of leaf (cm)\u003c/p\u003e \u003cp\u003eLeaf: Length of leaf petiole (cm)\u003c/p\u003e \u003cp\u003ePlant: Height at flowering (cm)\u003c/p\u003e \u003cp\u003eFlower: Length of flower petal (cm)\u003c/p\u003e \u003cp\u003eFlower: Width of flower petal (cm)\u003c/p\u003e \u003cp\u003eSiliqua: Length of siliqua (excluding beak) (mm)\u003c/p\u003e \u003cp\u003eSiliqua: Length of beak (mm)\u003c/p\u003e \u003cp\u003eSiliqua: Width of siliqua (mm)\u003c/p\u003e \u003cp\u003eSiliqua: Length of pedicel (mm)\u003c/p\u003e \u003cp\u003eSiliqua - Number of seeds in siliqua\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=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Model description\u003c/h2\u003e \u003cp\u003eIn this study, a Conv1D model, i.e., a one-dimensional convolutional neural network, was employed. Conv1D creates a convolution kernel that is convolved with the input data in one spatial dimension to generate results (Dewantara et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Batch Normalization was used in a model for classification purposes. This layer introduces random perturbation to the decision boundary of deep networks, forcing the model to learn boundaries with increased margins to the nearest training samples (Balestriero and Baraniuk, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). MaxPooling1D was used to perform pooling operations with the maximum values in a single spatial dimension when creating output data (Dewantara et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The flattening layer was utilized as it converts an array into a vector. The Dropout layer was applied as a method to overcome overfitting, which often occurs when using a deep learning approach. The Dropout layer randomly reduces the number of neuron units in the network to decrease the number of connections in each iteration of the learning process. The dense layer that was used, is a simple layer of neurons, where each neuron receives input data from all neurons in the previous layer (Dumane, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). The model utilized in this study is a sequential model.\u003c/p\u003e \u003cp\u003eThe sequential model applied to the analyzed data for mustard varieties is presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. It was initialized using the Keras library (an open-source Python library for machine learning). The layers of the model include two convolutional layers (Conv1D), two Batch Normalization layers, a pooling layer (MaxPooling1D), a flattening layer (Flatten), a Dropout layer, and two Dense layers. Each layer has its own parameters such as size, activation function, and input shape. In total, the model has 2,408 parameters, of which 2,344 are trainable, and 64 are not, as they are parameters of batch normalization layers. The activation function in layers 1, 3, and 7 is a rectified linear unit (ReLU), shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. If the input value is below zero, the output value is zero. When the input value rises above a certain threshold, it has a linear relationship with the dependent variable (Bisen, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). The formula for the ReLU function is provided below (Formula 1). The activation function in layer 9 is Softmax. This function is a combination of multiple sigmoid functions (Sharma, et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Since we know that the sigmoid function returns values in the range from 0 to 1, they can be treated as probabilities of data points for a specific class. Unlike sigmoid functions used for binary classification, the Softmax function can be used for multi-class classification problems. The function returns the probability for each data point for all individual classes. When building a network or model for multi-class classification, the output layer of the network will have the same number of neurons as the number of classes in the target object. The formula for the Softmax function is provided below (Formula 2).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\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\u003eModel summary (Model: \"sequential\").\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLayer Number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLayer\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eOutput Shape\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003econv1d\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 13, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ebatch_normalization\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 13, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003econv1d\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 11, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e784\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ebatch_normalization\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 11, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e64\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003emax_pooling1d\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 5, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eflatten\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003edense\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1296\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003edropout\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 16)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003edense_1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(None, 8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e136\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eTotal parameters\u003c/b\u003e:\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e2,408\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eTrainable parameters\u003c/b\u003e:\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e2,344\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cb\u003eNon-trainable params\u003c/b\u003e:\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003e64\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eFormula 1\u003c/b\u003e. ReLU activation function [Bisen, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2021\u003c/span\u003e].\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:f\\left(x\\right)=\\left\\{\\begin{array}{c}0,\\:\\:x\u0026lt;0\\\\\\:x,\\:\\:x\\ge\\:0\\end{array}\\right.$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere \u003cem\u003ex\u003c/em\u003e is the input to the ReLU activation function.\u003c/p\u003e \u003cp\u003e \u003cb\u003eFormula 2\u003c/b\u003e. Softmax activation function [Sharma, et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2020\u003c/span\u003e].\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:\\sigma\\:{\\left(z\\right)}_{j}=\\frac{{e}^{{z}_{j}}}{{\\sum\\:}_{k=1}^{K}{e}^{{z}_{k}}}\\:\\:\\:\\:for\\:j=1,\\dots\\:,K.$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWhere:\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:\\sigma\\:{\\left(z\\right)}_{j}\\)\u003c/span\u003e \u003c/span\u003e is the probability for every data point of all the individual classes,\u003c/p\u003e\u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:{\\left(z\\right)}_{j}\\)\u003c/span\u003e \u003c/span\u003e is the input to the Softmax activation function corresponding to the \u003cem\u003ej\u003c/em\u003e-th element.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eIn the model, applied loss function was 'categorical_crossentropy,' the optimizer was 'adam,' and the metric was 'accuracy.' The training parameters of the model were configured as follows: the 'epochs' parameter, determining the number of training epochs, is set to 80; the 'batch_size' parameter, defining the size of the training data batches, is set to 50; and applied 'validation_data' technique uses test data as the validation set during training.\u003c/p\u003e \u003cp\u003eThe results for the applied algorithm were not satisfactory, so cross-validation was applied but the results did not improve. It was then decided to use a new approach \u0026ndash; division into subsets. It turned out that the best solution is to divide it into 8 subsets as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The data was divided into eight subsets as presented in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. In the subset number 8 were eight varieties and seven varieties in each of the remaining subsets. The aim was to improve the classification accuracy of a complex dataset by creating smaller, more manageable subsets.\u003c/p\u003e \u003cp\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\u003eDivision of the variety dataset into subsets.\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\" colname=\"c1\"\u003e \u003cp\u003eSubset Number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVarieties\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1, 2, 3, 4, 5, 6, 7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8, 9, 11, 13, 14, 15, 16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e17, 18, 20, 21, 8000, 8001, 8023\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8002, 8003, 8006, 8008, 8009, 8010, 8012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8007, 8011, 8013, 8014, 8016, 8017, 8018\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19, 8019, 8020, 8021, 8022, 8024, 8025\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8026, 8027, 8028, 8029, 8030, 8031, 8032\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8033, 8034, 8035, 8036, 8037, 8038, 8039, 8040\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\u003eThe detailed operation of data subdivision for the data used in this study is presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. First, the user inputs the total number of subsets to be processed (variable \u003cem\u003en\u003c/em\u003e) which in our case is eight. Then, the loop counter \u003cem\u003ei\u003c/em\u003e is initialized to 1, indicating the start of the loop for processing each subset. For the current subset, a CNN model is applied. This step involves training the model using the subset data. Next, two types of plots are generated and shown. First plot is Train and Validation Accuracy which shows the accuracy of the model on both the training and validation data over epochs. Second plot is Confusion Matrix for the predictions, showing how well the model performed in classifying the data. Then, the evaluation metrics (e.g., accuracy, F1 score, precision, recall) for the current subset are printed. The classification decision for new data is printed, indicating whether the new data is classified as an existing variety or a new variety. A decision point checks if the loop counter \u003cem\u003ei\u003c/em\u003e equals \u003cem\u003en\u003c/em\u003e, indicating whether all subsets have been processed. If not, the loop counter \u003cem\u003ei\u003c/em\u003e is incremented by 1 and the process loops back to apply the CNN model to the next subset. If \u003cem\u003ei\u003c/em\u003e is equal to \u003cem\u003en\u003c/em\u003e, the process proceeds to the final step and the mean evaluation metrics for all subsets are printed and also the final classification decision based on the collective results of all subsets is printed.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe computations were conducted in Python 3.7.7, utilizing libraries such as tensorflow, scikit-learn, pandas, numpy, scipy, matplotlib, and seaborn.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Application of the model to the data\u003c/h2\u003e \u003cp\u003eSubsequently, the created Convolutional Neural Network (CNN) model in the Keras library was applied to each subset. The input data consists of 15 features describing a mustard variety, and the output label is an assigned variety number. Using the StandardScaler method from the sklearn.preprocessing library, the input data was standardized, ensuring that mean is equal to 0 and variance is equal to 1. This standardization process is achieved using the fit_transform() method, which calculates mean and standard deviation for each feature based on the available training data and then transforms the input data by subtracting the mean and dividing by the standard deviation. This process was applied to enhance the stability and effectiveness of machine learning.\u003c/p\u003e \u003cp\u003eSubsequently, using the \u0026lsquo;to_categorical\u0026rsquo; function from the tensorflow.keras.utils library, the output labels are transformed into binary form through a process known as one-hot encoding. This transformation ensures that each label is represented as a binary vector with a length equal to the number of classes, which is require for training a neural network.\u003c/p\u003e \u003cp\u003eThereafter, the data is divided into a training set and a test set using the \u0026lsquo;train_test_split\u0026rsquo; function from the sklearn.model_selection library. The parameter \u0026lsquo;test_size\u0026thinsp;=\u0026thinsp;0.2\u0026rsquo; indicates that 20% of the data will constitute the test set. The parameter \u0026lsquo;stratify\u0026thinsp;=\u0026thinsp;y\u0026rsquo; ensures that the split maintains the class proportions in the output labels. The parameter \u0026lsquo;random_state\u0026thinsp;=\u0026thinsp;42\u0026rsquo; sets the seed for randomness to ensure reproducibility of the split.\u003c/p\u003e \u003cp\u003eThe model is evaluated separately for each subset and in the end, mean is calculated for all evaluation metrics. Additionally, accuracy is calculated on the test set. The evaluation metrics include: train accuracy, test accuracy, F1 score, precision and recall.\u003c/p\u003e \u003cp\u003eThe model also assumes the possibility of the existence of previously unknown varieties. When providing the results of assigning the given data to a specific variety, the model may recognize these data as a new variety. Therefore, when submitting data to determine the variety, at least 10 observations should be provided. The model will iterate through all subsets, assign each observation to a specific variety, and then check if each observation is assigned to the same variety. If not, it means that the data is classified as a new variety. After checking this for all subsets, if in any of them, all given observations are classified into one specific variety, then the model classifies the given data as that variety. If in every subset the result is 'The new data is classified as a new variety.', it means that the data belongs to a new, previously unknown variety.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cp\u003eUsing the trained model, predictions are made on the new data. An experiment was conducted with data related to variety 8012 (10 observations randomly selected from 60 observations). These data were not used previously as training data. The results for each subset are presented in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. Train accuracy ranges from 0.905 to 0.975 and test accuracy ranges from 0.731 to 0.900. The subset where classification for variety 8012 was examined is subset 4. The decision in this subset is: 'The new data belongs to variety: 8012'.\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\u003eResults for input data of variety 8012.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSubset number\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTrain Accuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTest Accuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eF1 Score\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePrecision\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eRecall\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eDecision\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.956\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.731\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.734\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.742\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.731\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.975\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.840\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.845\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.906\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.793\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.806\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.789\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.970\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.899\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.905\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.900\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data belongs to variety: 8012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.943\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.774\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.775\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.788\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.774\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.905\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.881\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.876\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.881\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.945\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.854\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.851\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.957\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.854\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.933\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.825\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.847\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eThe new data is classified as a new variety.\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\u003eFor each subset, charts comparing train accuracy with validation accuracy were made. These charts are presented in Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. The model's accuracy varies depending on the number of epochs and regularization that was conducted using dropout layer. The charts show that there are no noteworthy differences between the compared accuracies, indicating that overfitting did not occur.\u003c/p\u003e \u003cp\u003eAdditionally, a confusion matrix was created for each subset, as shown in Table\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. It can be observed that the correct prediction ranges from 58\u0026ndash;100% (82% on average), with the highest accuracy in subset 4 ranging from 78\u0026ndash;100% (90% on average).\u003c/p\u003e \u003cp\u003eMean values for the evaluation metrics are presented in Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. Ultimately, the model classified the data as belonging to variety 8012. This indicates that the model is functioning correctly.\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\u003eMean values for the evaluation metrics and decision.\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\" colname=\"c1\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean Train Accuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.942\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean Accuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean F1 Score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean Precision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.833\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean Recall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.824\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDecision\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe new data belongs to variety: 8012\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eLee et al. (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) conducted a study on convolutional neural networks (CNN) to learn unsupervised feature representations for 44 different plant species. They concluded that learning features through CNN can provide superior feature representation for leaf images compared to hand-crafted features. Saini, et al. (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) presented a method for classifying healthy and diseased apple and grape leaves using a convolutional neural network. The network achieved a training accuracy of 99.96% and a testing accuracy of 99.90%. Kattenborn, et al. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) applied a segmentation approach based on CNN for detailed mapping of plant species communities. The demonstrated approach achieved segmentation and mapping accuracy of at least 84%. Wang, et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) constructed a multiscale CNN model with attention (AMSCNN) to enhance multiscale feature extraction capabilities. Proposed model was applied to plant species identification and achieved an accuracy of 73.62%. Ashturkar and Bhalchandra (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) proposed a designed CNN model and AlexNet model with application-oriented fine-tuning to classify plant species based on leaves. They achieved 91.19% accuracy for their CNN model and 97.41% accuracy for the AlexNet model. Wang and Su (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) compared CNN models used for detecting the phenotype of four cereal crops. They selected the latest CNNs related to organ counting, weed segmentation, biotic and abiotic stress assessment, and seed variety classification. Results showed the importance of neural network architecture, development strategy, and annotated datasets in designing models for different tasks, directly impacting CNN performance. Despite limited phenotypic information on seeds, the proposed CNN models were able to effectively classify seeds with an accuracy exceeding 95%. The authors also observed that deep CNN models have higher accuracy than shallow networks. Zhu et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) utilized deep CNN for plant species classification, testing existing classifiers and creating a new one for leaf, flower, fruit, stem, and branch classification. The lowest accuracy was obtained for stem classification, where the new classifier achieved the best result at 65.2% and existing classifiers ranged from 44.8\u0026ndash;58.7%. The highest accuracy was obtained for fruit classification, with accuracy ranging from 44.8\u0026ndash;66.3% for known CNN algorithms and 90.2% for the newly created model. Alom et al. (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) employed CNN for the classification of two types of B. Napus species, B. Rapa and B. Alba, considering flower, leaf, and packet features individually. The DenseNet201 model achieved the best results, getting 100% accuracy for flowers, 97% for leaves, and 100% for packets.\u003c/p\u003e \u003cp\u003eThe referenced studies focused on distinguishing between plant species, while this study specifically addresses the discrimination of plant varieties, which is more complex. The presence of a large number of varieties makes classification more challenging and in this study the number of varieties is 57. Morphological characteristics are unstable, they are created as a result of an environmental and climatic conditions, making phenotypic plasticity a consequence of adaptation (Korir et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Therefore, molecular markers, statistical tests, and software are employed for classification and characterization of plant varieties. As noted by Korir et al. (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), certain limitations create challenges to the development and application of modern scientific methods in identifying varieties of many cultivated plants. Taking all of this into consideration, the algorithm developed in this study classifies white mustard varieties very effectively and moreover, informs if a given data belongs to a previously unknown, new variety. Dividing data into subsets turned out to be an efficient method. Distinguishing 57 varieties at once is exceptionally difficult, so the implementation of a new idea, dividing the data into subsets, made this task easier. In studies on distinguishing varieties, the number of varieties in the dataset is usually much smaller than 57. Koklu, et al. (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) studied five rice varieties and achieved 100% classification success with CNN. Wang and Song (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) examined five varieties of sweet maize seeds using neural networks and achieved 95% accuracy.\u003c/p\u003e \u003cp\u003eWhile other studies typically dealt with datasets containing fewer varieties, this approach tackles a larger and more complex dataset. The developed algorithm can be widely used in agriculture, at all stages of cultivation. Thanks to this solution, varieties submitted for registration have the opportunity to achieve positive results in DUS tests and non-promising varieties could be eliminated from further research in terms of distinctness. In future studies, the proposed approach can be expanded by adding new varieties to the dataset and adjusting the parameters of the CNN architecture for these new varieties.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eL.S. did conceptualization, methodology, developed software and prepared all figures. B.Z. was responsible for data curation and supervision. L.S. wrote original draft and B.Z. reviewed and edited it.All authors approved the version to be published.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eData will be available on request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAbuassba AOM, Zhang D, Luo X, Shaheryar A, Ali H (2017) Improving Classification Performance through an Advanced Ensemble Based Heterogeneous Extreme Learning Machines. Comput Intell Neurosci. 10.1155/2017/3405463.\u003c/li\u003e\n\u003cli\u003eAlbawi S, Mohammed TA, Al-Zawi S (2017) Understanding of a convolutional neural network. 2017 International Conference on Engineering and Technology (ICET). 10.1109/icengtechnol.2017.8308186\u003c/li\u003e\n\u003cli\u003eAlferov A A, Chernova LS (2018) Influence of Fertilizers, Biomass of White Mustard, and Biopreparation Rhizoagrin on Yield and Grain Quality of Spring Wheat (\u003cem\u003eTriticum aestivum L.\u003c/em\u003e) and Sustainability of the Agroecosystem. Russian Agriculture Sciences 44: 546\u0026ndash;550. https://doi.org/10.3103/S1068367418060022\u003c/li\u003e\n\u003cli\u003eAlom M, Ali MY, Islam MT, Uddin AH, Rahman W (2023) Species classification of brassica napus based on flowers, leaves, and packets using deep neural networks. Journal of Agriculture and Food Research 14. https://doi.org/10.1016/j.jafr.2023.100658\u003c/li\u003e\n\u003cli\u003eAshturkar HH, Bhalchandra AS (2021) Plant Species Identification Using Convolutional Neural Network. Intelligent Data Communication Technologies and Internet of Things. Lecture Notes on Data Engineering and Communications Technologies. 10.1007/978-981-15-9509-7\u003c/li\u003e\n\u003cli\u003eAzizjon M, Jumabek A, Kim W (2020) 1D CNN based network intrusion detection with normalization on imbalanced data. 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC). 10.1109/icaiic48513.2020.9064976\u003c/li\u003e\n\u003cli\u003eBalestriero R, Baraniuk RG (2022) Batch normalization explained. arXiv. https://doi.org/10.48550/arXiv.2209.14778\u003c/li\u003e\n\u003cli\u003eBisen D (2021) Deep convolutional neural network based plant species recognition through features of leaf. Multimedia Tools and Applications 80:6443\u0026ndash;6456. https://doi.org/10.1007/s11042-020-10038-w\u003c/li\u003e\n\u003cli\u003eDewantara DS, Budi I, Ibrohim MO (2020) 3218IR at SemEval-2020 Task 11: Conv1D and Word Embedding in Propaganda Span Identification at News Articles. Proceedings of the 14th International Workshop on Semantic Evaluation 1716\u0026ndash;1721. 10.18653/v1/2020.semeval-1.225\u003c/li\u003e\n\u003cli\u003eDumane G (2020) Introduction to Convolutional Neural Network (CNN) using Tensorflow. Towards Data Science. https://towardsdatascience.com/introduction-to-convolutional-neural-network-cnn-de73f69c5b83 [available: 13.11.2023]\u003c/li\u003e\n\u003cli\u003eGrygier A (2023) Mustard Seeds as a Bioactive Component of Food. Food Reviews International 39(7):4088-4101. 10.1080/87559129.2021.2015774\u003c/li\u003e\n\u003cli\u003eGuo X, Lu H, Lin Y, Chen B, Wu C, Cui Z, Wang Y, Xu Y (2013) Skin penetration of topically applied white mustard extract and its effects on epidermal Langerhans cells and cytokines. International Journal of Pharmaceutics 457(1):136-142. 10.1016/j.ijpharm.2013.09.015.\u003c/li\u003e\n\u003cli\u003eKattenborn T, Eichel J, Fassnacht FE (2019) Convolutional Neural Networks enable efficient, accurate and fine-grained segmentation of plant species and communities from high-resolution UAV imagery. Scientific Reports 9(17656). https://doi.org/10.1038/s41598-019-53797-9\u003c/li\u003e\n\u003cli\u003eKoklu M, Cinar I, Taspinar YS (2021) Classification of rice varieties with deep learning methods. Computers and Electronics in Agriculture 187. https://doi.org/10.1016/j.compag.2021.106285.\u003c/li\u003e\n\u003cli\u003eKorir NK, Han J, Shangguan L, Wang C, Kayesh E, Zhang Y, Fang J (2013) Plant variety and cultivar identification: advances and prospects. Critical Reviews in Biotechnology 33(2):111\u0026ndash;125. 10.3109/07388551.2012.675314\u003c/li\u003e\n\u003cli\u003eLee SH, Chan CS, Wilkin P, Remagnino P (2015) Deep-plant: Plant identification with convolutional neural networks. 2015 IEEE International Conference on Image Processing (ICIP). 10.1109/icip.2015.7350839\u003c/li\u003e\n\u003cli\u003eLi Z, Liu F, Yang W, Peng S, Zhou J (2021) A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects. IEEE Transactions on Neural Networks and Learning Systems: 1\u0026ndash;21. 10.1109/tnnls.2021.3084827\u003c/li\u003e\n\u003cli\u003eSaini G, Khamparia A, Luhach AK (2020) Classification of Plants Using Convolutional Neural Network. First International Conference on Sustainable Technologies for Computational Intelligence. Advances in Intelligent Systems and Computing. 10.1007/978-981-15-0029-9\u003c/li\u003e\n\u003cli\u003eSharma JR (2006) Statistical and Biometrical Techniques in Plant Breeding. New Age International, New Delhi.\u003c/li\u003e\n\u003cli\u003eSharma S, Sharma S, Athaiya A (2020) Activation functions in neural networks. International Journal of Engineering Applied Sciences and Technology 4(12):310\u0026ndash;316.\u003c/li\u003e\n\u003cli\u003eWang X, Zhang C, Zhang S (2021) Multiscale Convolutional Neural Networks with Attention for Plant Species Recognition. Computational Intelligence and Neuroscience. 10.1155/2021/5529905\u003c/li\u003e\n\u003cli\u003eWang Y, Song S (2023) Variety identification of sweet maize seeds based on hyperspectral imaging combined with deep learning. Infrared Physics \u0026amp; Technology 130. https://doi.org/10.1016/j.infrared.2023.104611.\u003c/li\u003e\n\u003cli\u003eWang Y-H, Su W-H (2022) Convolutional Neural Networks in Computer Vision for Grain Crop Phenotyping: A Review. Agronomy 12(11). https://doi.org/10.3390/agronomy12112659\u003c/li\u003e\n\u003cli\u003eZhang Z, Zhao Y, Liao X, Shi W, Li K, Zou Q, Peng S (2018) Deep learning in omics: a survey and guideline. Briefings in Functional Genomics. 10.1093/bfgp/ely030\u003c/li\u003e\n\u003cli\u003eZhu H, Liu Q, Qi Y, Huang X, Jiang F, Zhang S (2018) Plant identification based on very deep convolutional neural networks. Multimedia Tools and Applications 7. 10.1007/s11042-017-5578-9\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 5 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":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"euphytica","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"euph","sideBox":"Learn more about [Euphytica](https://www.springer.com/journal/10681)","snPcode":"10681","submissionUrl":"https://submission.springernature.com/new-submission/10681/3","title":"Euphytica","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Agricultural Algorithms, Classification, Convolutional Neural Networks, Mustard, Variety Recognition","lastPublishedDoi":"10.21203/rs.3.rs-4752375/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4752375/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe aim of this study was to develop and apply a Convolutional Neural Network (CNN) model to recognize and classify white mustard (\u003cem\u003eSinapis alba L.\u003c/em\u003e) varieties, addressing the complex task of discriminating among 57 varieties. Utilizing a one-dimensional CNN model, the research focused on multivariate analysis based on a set of 15 traits. The CNN architecture included convolutional layers, batch normalization, pooling, flattening, dropout, and dense layers. The model demonstrated effectiveness in classifying varieties, achieving high accuracy and providing valuable insights into potential new varieties. Subset division, a new approach, was applied. Evaluation metrics, including accuracy, F1 score, precision, and recall, were calculated for eight subsets, confirming the model's robust performance. Comparison with existing studies highlights the uniqueness of the proposed algorithm in addressing variety discrimination. The approach contributes to agricultural advancements, offering a reliable tool for breeders to assess variety distinctness and streamline the testing process. The model's ability to detect unknown varieties further enhances its utility in agricultural research covering a comprehensive and impactful advancement in variety classification.\u003c/p\u003e","manuscriptTitle":"Applying Convolutional Neural Networks for Mustard Variety Recognition","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-08-09 08:23:06","doi":"10.21203/rs.3.rs-4752375/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2024-12-20T15:55:50+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-12T19:59:33+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-12T16:12:31+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-12T03:02:55+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2024-12-11T09:40:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"17153477169102545908091598160023813170","date":"2024-11-29T08:29:59+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"238443654729961309320756080734699579406","date":"2024-11-28T20:35:10+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"152880827037222306247387885470816021551","date":"2024-11-28T12:38:38+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"243165810916404738454902257256092759119","date":"2024-11-28T12:37:25+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2024-07-20T12:02:20+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2024-07-17T03:25:42+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2024-07-17T03:25:09+00:00","index":"","fulltext":""},{"type":"submitted","content":"Euphytica","date":"2024-07-16T21:11:38+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"euphytica","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"euph","sideBox":"Learn more about [Euphytica](https://www.springer.com/journal/10681)","snPcode":"10681","submissionUrl":"https://submission.springernature.com/new-submission/10681/3","title":"Euphytica","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"0cc50533-afaf-4920-865f-5927d3074de3","owner":[],"postedDate":"August 9th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-01-20T16:08:48+00:00","versionOfRecord":{"articleIdentity":"rs-4752375","link":"https://doi.org/10.1007/s10681-025-03461-3","journal":{"identity":"euphytica","isVorOnly":false,"title":"Euphytica"},"publishedOn":"2025-01-16 15:57:19","publishedOnDateReadable":"January 16th, 2025"},"versionCreatedAt":"2024-08-09 08:23:06","video":"","vorDoi":"10.1007/s10681-025-03461-3","vorDoiUrl":"https://doi.org/10.1007/s10681-025-03461-3","workflowStages":[]},"version":"v1","identity":"rs-4752375","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4752375","identity":"rs-4752375","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}