Centralized Training and Decentralized Invocation: An Innovative Intelligent Recognition and Pricing System for Retail Products

Centralized Training and Decentralized Invocation: An Innovative Intelligent Recognition and Pricing System for Retail Products | 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 Centralized Training and Decentralized Invocation: An Innovative Intelligent Recognition and Pricing System for Retail Products Yunfei Wang, Yaping Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8297422/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract In the retail industry, efficient product recognition and pricing systems are crucial for enhancing customer experience and operational efficiency. Traditional methods, which often rely on manual input or image selection, are prone to inefficiency and errors. This paper introduces an innovative intelligent system that integrates cameras, electronic scales, and thermal printers with cloud-based deep learning technologies. The core of this approach lies in centralized training of a convolutional neural network (CNN) model on a high-performance cloud platform, enabling decentralized invocation of the recognition results at retail terminals. By uploading product images to the cloud, the system achieves rapid and accurate recognition, with precision, recall, and F1-Score exceeding 95.1%, 95.0%, and 96.2%, respectively. We demonstrate that this architecture not only optimizes the shopping process but also facilitates the large-scale application of artificial intelligence in retail environments. The system's feasibility and efficiency have been validated through extensive testing on 30 types of fruits and vegetables, indicating its strong potential for widespread adoption in supermarkets and agricultural markets. Integration Cloud Deep learning Recognition Weight Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1 Introduction In the retail sector, conventional product recognition and pricing methods often depend on manual entry of product codes or selection from image libraries. These approaches are inefficient, susceptible to errors, and can lead to increased customer wait times and higher operational costs for supermarkets. Furthermore, unmanned stores and self-checkout systems face limitations in recognizing pre-packaged products, particularly in identifying spoilage that is not visible from outside the package. Therefore, developing an intelligent system for accurate and efficient retail product recognition is of paramount importance. Prior research has explored various techniques. Google and Amazon's recognition services have been utilized for object detection, but these typically do not integrate weighing functionality [1,2]. Ni Yunfeng developed an image acquisition system using cameras and light sources, employing MATLAB to differentiate apples and pears based on color features [3]. Tao Weihua created image acquisition software on a Cortex A8 board to extract color and texture features for fruit and vegetable classification [4]. Yao Xinguang designed a fruit recognition and weighing system using an OPENMV module, identifying fruits through hierarchical scanning of color, size, and similarity [5]. Wang Hui enhanced DarkNet-53 model with YOLOv3 algorithm to improve fruit recognition accuracy [6]. Zou Bin optimized GoogleNet and ResNet-50 structures to develop a fruit recognition model that increased accuracy and reduced processing time [7]. Jin Mei designed an FPGA-based system to meet real-time recognition requirements [8]. Wang Yunfei developed an intelligent electronic scale with an image processing module to identify fruit and vegetable codes for pricing [9–11]. Zhang Lulu designed a self-service recognition system, replacing the YOLOv5 backbone with a lighter network to boost recognition speed [12]. Chen Dengfeng used pressure sensors and faster R-CNN algorithm, combining image recognition with weight data to generate settlement barcodes for automated checkout [13]. Convolutional Neural Network (CNN)-based intelligent fruits classification utilizing the bilinear pooling with heterogeneous streams is proposed [14]. A common limitation in prior research is that models are typically trained and invoked locally on specific devices. This approach hinders sharing and deployment across different retail terminals, thereby constraining the scope and pace of intelligent upgrades in the industry. This paper proposes a paradigm shift: leveraging high-performance cloud computing resources to centrally train deep learning models on large datasets of product images. This process generates models with high recognition precision and recall. In practice, retail terminals can then easily invoke these pre-trained models from the cloud for product recognition, eliminating the need for deploying and maintaining extensive training infrastructure at each terminal. This cloud-centric strategy effectively removes the barriers to the broad and rapid intelligent upgrading of retail systems. 2 Method 2.1 System architecture The proposed system, based on the principle of centralized training and decentralized invocation of deep learning models, integrates computer vision, deep learning, serial communication, thermal printing, and weighing technologies. The overall architecture, depicted in Fig. 1 , consists of the following layers: Hardware layer: This includes a pressure sensor (connected to a microcontroller via an HX711AD module), a high-definition camera, a thermal printer, and a computer acting as the control center. The pressure sensor measures weight and transmits data to the computer via serial communication. The camera captures images of products for recognition. The thermal printer generates receipts containing product information. Data transmission layer: This layer facilitates data exchange between hardware components through serial ports (USB or RS232). Weight data from the sensor is sent to the microcontroller, which relays it to the computer. The computer encodes images captured by the camera and uploads them to the cloud, subsequently receiving recognition results back from the cloud. Cloud-based deep learning model: A Convolutional Neural Network (CNN) model is developed and trained on a cloud AI platform (e.g., Baidu AI Studio) to recognize product types. The input is the product image uploaded from the retail terminal, and the output is the product classification. Software layer: An integrated control system, developed in C#, manages camera capture, cloud model invocation, weight data reception, amount calculation, barcode generation, and thermal printer control. This layer also handles serial communication with the microcontroller and pressure sensor. User interaction layer: When a customer places fruits or vegetables on the electronic scale, the system automatically executes recognition, weighing, pricing, and prints a receipt. This fully automated process significantly enhances shopping efficiency and customer experience. 2.2 Technical route The technical workflow to achieve product recognition, weighing, pricing, and receipt printing encompasses the following steps: Model construction and training: A CNN model is constructed and trained on a cloud-based deep learning platform using a prepared training set of images. Model parameters are optimized during this phase. Software development: An integrated control system is developed in C# to enable communication with hardware devices, invocation of the cloud model, data processing, and printing functions. Hardware integration: Electronic barcode scales, cameras, and thermal printers are connected to the computer via USB or serial ports. Necessary configurations and debugging are performed to ensure seamless integration and communication. System integration and testing: Comprehensive joint debugging of hardware and software is conducted to ensure all components work harmoniously to complete the required tasks. The system's feasibility, precision, and recall are rigorously verified. 2.3 Deep learning model 2.3.1 Model establishment on cloud platform To achieve efficient recognition, a CNN model was established and trained on the Baidu AI Open Platform. High-performance server resources, including Tesla V100 16GB GPUs, 12-core CPUs, and 56GB of RAM, were leveraged to ensure training efficiency. The architecture of the constructed CNN is illustrated in Fig. 2 . The network comprises an input layer, six convolutional layers, a fully connected layer, and a dropout layer. To enhance generalization capability, max-pooling layers are inserted between the last convolutional layer and the fully connected layer. The hidden layers employ the Maxout activation function. The input layer starts with 64 channels, and this number is doubled at each subsequent layer until reaching 512. A flatten function then converts the data into a one-dimensional vector. This vector passes through a 1×1×512 fully connected layer. To mitigate overfitting, a dropout layer is applied. The final classification result is output by a softmax function. 2.3.2 Dataset preparation To ensure model efficacy and robustness, a dataset of 30 types of fruits and vegetables was curated. Images were captured using cameras mounted on the barcode scales, covering single or multiple items, various stacking configurations, and products inside or outside shopping bags. For each type, 200 images were collected. These were randomly split into 120 for training, 40 for validation, and 40 for testing, as detailed in Table 1 . Sample images from the training set are shown in Fig. 3 . Table 1 Composition of training, validation, and test sets Quantity Requires Tagging? Requires Internet? Type 30 / / Training 120 per type Yes No Validation 40 per type No No Test 40 per type No Yes Total 6000 / / 2.3.3 Model training and validation Training and validation are critical for ensuring the model's precision and recall, forming the foundation for reliable product recognition and pricing. The process is outlined in Fig. 4 . During the training phase, the tagged training set (3600 images across 30 categories) was fed into the CNN model. The training involves iterative forward propagation to generate predictions and backpropagation with a gradient descent algorithm to update model parameters, minimizing the error between predictions and actual labels. Data augmentation was applied using the Baidu EasyDL platform's default configuration. Hyperparameters were selected via the platform's automatic search feature, and the number of training epochs was set according to its automatic configuration. During the validation phase, the untagged validation set (1200 images) was fed into the trained model to evaluate its performance using precision, recall, and the F1-Score, defined as: P \(\:\text{r}\text{e}\text{c}\text{i}\text{s}\text{i}\text{o}\text{n}=\frac{\text{T}\text{P}}{\text{T}\text{P}+\text{F}\text{P}}\) (1) $$\:\text{R}\text{e}\text{c}\text{a}\text{l}\text{l}=\frac{\text{T}\text{P}}{\text{T}\text{P}+\text{F}\text{N}}$$ 2 $$\:{\text{F}}_{1}-\text{S}\text{c}\text{o}\text{r}\text{e}\:=\frac{2\times\:\text{P}\text{r}\text{e}\text{c}\text{i}\text{s}\text{i}\text{o}\text{n}\times\:\text{R}\text{e}\text{c}\text{a}\text{l}\text{l}}{\text{P}\text{r}\text{e}\text{c}\text{i}\text{s}\text{i}\text{o}\text{n}+\text{R}\text{e}\text{c}\text{a}\text{l}\text{l}}$$ 3 Where: TP (True Positive) is the number of positive samples correctly identified. FP (False Positive) is the number of negative samples incorrectly identified as positive. FN (False Negative) is the number of positive samples incorrectly identified as negative. A validation F1-Score threshold of ≥ 99.5% was set. If the model failed to meet this threshold, it underwent further training with hyperparameter adjustments until the target was achieved. 2.4 System integration 2.4.1 Weight data acquisition The pressure sensor in the electronic scale uses an HX711AD module to convert analog weight signals into digital data. This data is transmitted to the computer via serial communication. A C# program was developed to receive and decode the weight data from the serial port. 2.4.2 Image capture A camera is positioned above the electronic scale. When the load sensor detects an object and transmits weight data to the computer, it simultaneously triggers the camera to capture an image. This image is then sent to the computer for encoding before being submitted to the cloud model for recognition. 2.4.3 Integration with cloud model Upon receiving an image from the camera, the C# program encodes it into Base64 format. It then invokes the cloud-based deep learning model via an API call over the internet. The cloud model processes the image and returns the recognition result to the retail terminal computer. 2.4.4 Amount calculation and barcode generation After obtaining the recognition result and weight data, the C# program retrieves the corresponding unit price from a preset database and calculates the total cost. An EAN-128 barcode is generated, encoding product ID, category name, price, and weight using a standard encoding algorithm. 2.4.5 receipt printing The C# program sends the product details (type name, unit price, weight, total cost) and the generated barcode to the thermal printer. Control code ensures the barcode and text are printed clearly and without errors. Through careful hardware selection and software design, the electronic scale, camera, and thermal printer are seamlessly integrated into a single terminal device, as shown in Fig. 5 . 2.5 Testing experiments During the testing phase, the entire system was evaluated using the 1200-image test set. The tests simulated real-world usage, requiring internet connectivity for the terminal device to communicate with the cloud model. The main steps were: 1) Simulating a customer placing a fruit or vegetable on the weighing sensor (Fig. 6); 2) Receiving weight data from the scale via serial communication; 3) Triggering the camera to capture an image of the item (one from the test set); 4) Uploading the encoded image to the cloud for recognition by the deep learning model; 5) Receiving the product type information from the cloud; 6) Retrieving the unit price based on the recognized product code; 7) Calculating the total amount based on unit price and weight; 8) Generating an electronic barcode; 9) Printing a receipt with the barcode (Fig. 7); 10) Removing the product from the sensor to prepare for the next customer. This process is fully automated, requiring manual intervention only for placing and removing product. A sample printed receipt is shown in Fig. 7. 3 Results A comparison between this research and prior work [1,2] is presented in Table 2 , highlighting key differences in model training location, invocation location, and scalability. After 1200 tests on 30 different types of fruits and vegetables, where the retail terminal invoked the pre-trained cloud model, the system achieved high recognition precision (95.1%), recall (95.0%), and F1-Score (96.2%). Detailed results for each type are listed in Table 3 . The heatmap of the confusion matrix is shown in Fig. 8 . The serial numbers on the axes of Fig. 8 correspond to the serial numbers in the first column of Table 3 . Most categories achieved perfect scores. However, cherries and grapes showed a slight probability of being misclassified for one another. A similar confusion was observed between apples and dangshan pears, and between Hami melons and watermelons. Table 2 Comparison with prior research Model training location Model invocation location Readily deployable to multiple terminals? Prior research Local Local No This research Cloud Decentralized terminal Yes 4 Discussion The results demonstrate the feasibility of the proposed architecture based on centralized cloud training and decentralized invocation for product recognition. Compared to previous studies where both training and recognition were conducted locally on retail terminals [12,13], this approach significantly lowers the barrier to deploying intelligent recognition systems. Electronic barcode scales in various settings—supermarkets or agricultural markets—can now leverage advanced AI recognition as long as a basic internet connection is available. Notably, the system exhibits excellent scalability. Introducing a new product type only requires training the model on this specific type. Once validated, the updated model can be deployed to all terminals, immediately making the new product recognizable across the entire network. Grape and cherry are both small in round shape and similar in color, which makes it difficult to completely distinguish between them. Additionally, apple and dangshan pear share a medium-sized, round form with comparable colors and textures, while watermelon and hami melon are both large and spherical, featuring green outer skins. This imposes a constraint on improvement of precision and recall. The primary challenges to improving precision and recall stem from visual similarities between certain fruits. Grapes and cherries are both small and round with similar colors. Apples and dangshan pears share a medium-sized round form with comparable colors and textures. Watermelons and hami melons are both large and spherical with green rinds. These inherent visual similarities present the main constraint on further precision and recall improvements. In comparison with specific application scenarios merely focusing on recognition [15,16] or classification [17,18], this research realizes the integration of classification and weighing, thereby achieving a more universal scope of application. The intelligent recognition and pricing system, with its cloud-centric design, is readily deployable across multiple retail terminals. It is particularly suitable for chain supermarkets, where unified standards for product types, barcode scales, and environmental lighting can be enforced across locations. By consistently uploading images to a centrally maintained and improved cloud model, high levels of precision and recall can be sustained and enhanced over time. Table 3 Recognition results on the test set No. Type Test frequency Avg. weight(kg) Precision Recall F1-Score 1 Jujube 40 9.713 100% 100% 100% 2 Pitaya 40 9.268 100% 100% 100% 3 Cherry 40 8.975 97.4% 95.0% 96.2% 4 Pineapple 40 8.741 100% 100% 100% 5 Onion 40 8.635 100% 100% 100% 6 Kefuli 40 7.515 100% 100% 100% 7 Tomatoes 40 7.400 100% 100% 100% 8 Peach 40 6.999 97.5% 97.5% 97.5% 9 Pumpkin 40 6.782 100% 100% 100% 10 Potato 40 6.131 100% 100% 100% 11 Jackfruit 40 5.236 100% 100% 100% 12 Knight Orange 40 5.177 100% 100% 100% 13 Durian 40 3.915 100% 100% 100% 14 Water-melon 40 3.144 97.5% 97.5% 97.5% 15 Dangshan Pear 40 3.604 97.6% 100% 98.8% 16 Grape 40 3.076 95.1% 97.5% 96.3% 17 Apple 40 2.972 97.4% 95.0% 96.2% 18 Persim-mon 40 2.964 100% 100% 100% 19 Green pepper 40 2.566 100% 100% 100% 20 Hami melon 40 2.457 97.5% 97.5% 97.5% 21 Mango 40 2.432 100% 100% 100% 22 Papaya 40 2.138 100% 100% 100% 23 Banana 40 2.097 100% 100% 100% 24 Kiwifruit 40 2.068 100% 100% 100% 25 Straw-berry 40 1.979 100% 100% 100% 26 Coconut 40 1.874 100% 100% 100% 27 Fig 40 1.649 100% 100% 100% 28 Grapefruit 40 1.560 100% 100% 100% 29 Avocado 40 0.852 100% 100% 100% 30 Blueberry 40 0.476 100% 100% 100% This study successfully demonstrates the principle of centralized training and decentralized invocation for fruit and vegetable recognition. However, it also highlights areas for future investigation. Potential issues related to network latency and data security were not covered in this work. Network delays could impact recognition speed during peak hours, potentially reducing system responsiveness. Furthermore, uploading product images to the cloud necessitates robust data privacy and security measures to protect potentially sensitive information. Investigating methods to reduce latency and enhance data security will be key directions for our future research. 5 Conclusion This research has successfully implemented a centralized training and decentralized invocation framework for fruit and vegetable recognition by establishing a deep learning model on a cloud platform. At the decentralized retail terminals, functionalities for image encoding, uploading, receiving recognition results, and printing settlement barcodes were developed and integrated. The deployment of this system across multiple terminals represents an innovative integration of artificial intelligence with the retail industry, paving the way for the widespread and scalable application of artificial intelligence in this domain. Declarations Author Contribution Yunfei Wang wrote the main manuscript text and Yaping Li prepared figures 1-8. All authors reviewed the manuscript. Acknowledgements This research is supported by science and technology project of Henan Province, China (No.162102210320). References Fu Yuesheng, Song Jian, Xie Fuxiang.: Circular Fruit and Vegetable Classification Based on Optimized GoogLeNet, IEEE Access. 9, 113599–113611(2021) Sharma, V.: Object Detection and Recogni-tion using Amazon Rekognition with Boto3, 6th international conference on trends in electronics and informatics, Tirunelveli, India.727–732 (2022) Ni Yunfeng, Ye Jian, Fan Jiaojiao.: Fruit sorting system based on image Recognition, Jiangsu Agricultural Sciences. 49 (1), 170–176 (2021) Tao Weihua, Zhao Li, Xi Ji.: Fruits and vegetables recognition based on color and texture features, Transactions of Chinese Society of Agricultural Engineering. 30(16), 305–311 (2014) Yao Xinguang, Su Xiulan, Li Qinyu.: Design of Intelligent Fruit Recognition and Weighing System, Electronics World. 43(1), 174–175(2021) Wang Hui, Zhang Fan, Liu Xiaofeng.: Fruit image recognition based on DarkNet-53 and YOLOv3, Journal of Northeast Normal Uiversity (Natural Science Edition). 52(4), 60–65(2020) Zou Bin.: Vegetable-Fruit Recognition and Price Estimating Using Neural Networks, M.A. dissertation, Dept. School of Electronics and Information, Harbin Institute of Technology, Haerbin, Hei Longjiang, China (2023) Jin Mei, Zeng Xin, Zhang Liguo.: A fruit identification system design based on field programmable gate array, Chinese High Technology. 34(6), 616–623(2024) Wang Yunfei.: An electronic barcode scale intelligent recognition of bulk commodity item numbers, China patent 2019211701592 (2019) Wang Yunfei.: A smart membership fingerprint electronic scale for identifying products, China patent 2019216994458 (2019) Wang Yunfei.: A smart membership card electronic scale for identifying products, China patent 2019215705817 (2019) Zhang Lulu.: Design of an automatic recognition system for fruits and vegetables based on machine vision, Qingdao University, M.A. dissertation, Dept. School of automation, Qingdao University, Qingdao, Shan Dong, China (2023) Chen Dengfeng, Zhou Yao, Duan You.: Intelligent fruit and vegetable settlement system based on computer vision, Informatization Research. 45(2) 65–70 (2019) Prakash, Achanta Jyothi, Prakasam, P.: An intelligent fruits classification in precision agriculture using bilinear pooling convolutional neural networks, The Visual Computer. 39(5) 1–17(2022) Huixian Lin, Haidong Deng, Hong Du.: Low visibility underwater biological target detection based on the improved YOLOV5s, The Visual Computer. 41(12) 1–14(2025) Anam Nazir, Muhammad Nadeem Cheema, Bin Sheng.: OFF-eNET: An Optimally Fused Fully End-to-End Network for Automatic Dense Volumetric 3D Intracranial Blood Vessels Segmentation, IEEE Transactions on Image Processing. 29 7192–7202 (2020). Yang Wen, Ying Zeng, Shuang Liu.: Dsf-net: a dual-stream fusion network integrating structural and detailed features for fundus-based diabetic retinopathy classification, The Visual Computer. 41(15) 1–13(2025). Masood Anum, Sheng Bin, Yang Po.: Automated Decision Support System for Lung Cancer Detection and Classification via Enhanced RFCN with Multilayer Fusion RPN, IEEE Transactions on Industrial Informatics. 46(12) 7791–7801(2020). Additional Declarations No competing interests reported. Supplementary Files Authors.docx Cite Share Download PDF Status: Posted Version 1 posted 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-8297422","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":556903606,"identity":"1bcf7020-aa9d-47aa-bec6-f9a41e0332bc","order_by":0,"name":"Yunfei Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5UlEQVRIiWNgGAWjYFACxgYgYcPPwMx+8EFCRQ0xWphBWtIkG5h5kg0enDlGjBZ2kJbDkkDSTPJhCzNhDQbHD7ZJ8+5gluBnZ0irSGxgY+Bv707Ar+VMIlDLGTYJyWbGYzcSd8gwSJw5uwGvFrMDIC1tPHUGhxnSbiSeYWMwkMgloOX8Q5AWCQn7wwxmBYltzERouQG2xUDCgJnBjIEoLfY3HjZbzm1LkJA4zJMskXDmGA9Bv0j2pz+88bbtvwR///GDH39U1Mjxt/fi1wIELBLIPB5CykGA+QMxqkbBKBgFo2AEAwDlIkeKka3ZAgAAAABJRU5ErkJggg==","orcid":"","institution":"Henan Mechanical \u0026 Electrical Vocational College","correspondingAuthor":true,"prefix":"","firstName":"Yunfei","middleName":"","lastName":"Wang","suffix":""},{"id":556903607,"identity":"2d75db22-9f9b-4393-b1ae-d6443dab929b","order_by":1,"name":"Yaping Li","email":"","orcid":"","institution":"Henan Mechanical \u0026 Electrical Vocational College","correspondingAuthor":false,"prefix":"","firstName":"Yaping","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2025-12-07 03:08:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8297422/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8297422/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97795356,"identity":"431659c0-0cb8-46fe-a49e-ec83c256bd8e","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1143844,"visible":true,"origin":"","legend":"","description":"","filename":"CentralizedTrainingandDecentralizedInvocationAnInnovativeIntelligentRecognitionandPricingSystemforRetailProducts20251129.docx","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/5810d12764af5d4b0711d8f0.docx"},{"id":97795272,"identity":"e49db241-36dc-4f0f-8798-005f0f90ef8f","added_by":"auto","created_at":"2025-12-09 12:38:43","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3809,"visible":true,"origin":"","legend":"","description":"","filename":"7ea727ec44ae4dc698397f0638af01b5.json","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/7505d4ec99111dc49aa2adc6.json"},{"id":97795325,"identity":"260797ad-3d12-4e6b-b25d-0e462bb744ca","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":74462,"visible":true,"origin":"","legend":"","description":"","filename":"7ea727ec44ae4dc698397f0638af01b51enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/926f8636adc5079b292f075c.xml"},{"id":97795364,"identity":"000e4fb0-4f55-4843-97c4-c8faba834020","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70123,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/3b15df78d00f9a7f63202136.png"},{"id":97795304,"identity":"15029794-312f-4f50-a8b9-262d29547fa5","added_by":"auto","created_at":"2025-12-09 12:38:44","extension":"jpeg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7207,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage10.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/8ff44eb2c6bb1df4926936fb.jpeg"},{"id":97795369,"identity":"8adff920-69a5-49f6-ac1b-129dc6cec6f3","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":98475,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/d0d8300d5d69f636edab9c90.png"},{"id":97795313,"identity":"ee5728cc-f42d-44e8-aee7-f7366caebdab","added_by":"auto","created_at":"2025-12-09 12:38:45","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":355204,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/cc99ae9844ad446853f3ceeb.png"},{"id":97795350,"identity":"505ad4a8-089b-435d-a5c6-82ab83d96477","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":25291,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/aa7842618dd1b0c2e0d2d21e.png"},{"id":97795308,"identity":"075f401a-e6ed-4405-b1ec-5250ffd16d09","added_by":"auto","created_at":"2025-12-09 12:38:45","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":35650,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/ee621b65fd1470ca99b40599.jpeg"},{"id":97795363,"identity":"afa20fd4-1756-469a-ba5b-0b05bcdb702d","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":267948,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/e2d899b3cce08f242a1444cc.png"},{"id":97795367,"identity":"86067582-70e5-48dc-bde9-0fd391e3d5cf","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":207823,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/60af6a60f9721c2abd0dbb68.png"},{"id":97795362,"identity":"793246b5-8c24-4058-b671-3333674819ae","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"jpeg","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24549,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/f6a163da76e5c7e51961c2f2.jpeg"},{"id":97795359,"identity":"75b4129b-f856-48ef-b5df-b6863db8294f","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"jpeg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7460,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/c5ca3b93022cfa1b93fcc8e2.jpeg"},{"id":97897474,"identity":"3b5dd742-9585-4a78-a244-d1c1fa72867f","added_by":"auto","created_at":"2025-12-10 15:37:51","extension":"wmf","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":184,"visible":true,"origin":"","legend":"","description":"","filename":"image4.wmf","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/b65a3cae8913bf15e437f77d.wmf"},{"id":97795374,"identity":"0e63855d-c1d8-46de-8a43-d688cf9daea0","added_by":"auto","created_at":"2025-12-09 12:38:48","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":24662,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/4d36831fe27be8c0d97ec96a.png"},{"id":97795334,"identity":"9ada72c9-30df-4b25-bb8f-c4ae7ba2be22","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17037,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage10.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/b52e5cae3f2ff665f0c7d767.png"},{"id":97795365,"identity":"c3fbc2cb-5ed4-4495-aeb5-820fd48d6252","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10187,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/4150ec10b51c988a36eca764.png"},{"id":97795352,"identity":"ae7ec5ab-dfbc-4c8c-ae94-f13de435a92b","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":64258,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/8553e5db9c93106bcca0c2e6.png"},{"id":97897403,"identity":"1e17a5e0-eacc-464b-a735-6e3d6a85d44d","added_by":"auto","created_at":"2025-12-10 15:37:48","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4354,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/8118be8dbf67fc0b4c469b34.png"},{"id":97795358,"identity":"8ca043c8-5d8b-4710-b4bb-7d117e0e4a82","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50407,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/81853a5d802d02d37e6dbac7.png"},{"id":97795338,"identity":"8ab7701a-52b4-45ce-afbb-d8dc68da0805","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":39630,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/79e59e0db67f7a9882ebe36a.png"},{"id":97795329,"identity":"3b3cda0e-3a20-467d-b6f9-ee2e24cf4ba9","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":27164,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/c822f0626fb78537fb8ba8ea.png"},{"id":97898745,"identity":"774cc1ec-5e98-4f47-a917-64cada18d85d","added_by":"auto","created_at":"2025-12-10 15:39:33","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17055,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/b2b58b92eb8184b6b799c047.png"},{"id":97795373,"identity":"4e5b8e4b-129a-4b71-9082-bf710c4b82db","added_by":"auto","created_at":"2025-12-09 12:38:48","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11916,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/c081fe945d2ffcd4fde03846.png"},{"id":97795366,"identity":"3df94ef8-037d-4c20-9d31-b1a951a9a87a","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":135,"visible":true,"origin":"","legend":"","description":"","filename":"Onlineimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/44f01b167ebe5ab0e0ebca5b.png"},{"id":97795375,"identity":"9962e454-4da5-4f35-924a-734634daa5af","added_by":"auto","created_at":"2025-12-09 12:38:48","extension":"xml","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":71298,"visible":true,"origin":"","legend":"","description":"","filename":"7ea727ec44ae4dc698397f0638af01b51structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/c04f62942235e07cfc070705.xml"},{"id":97795370,"identity":"fc19bb01-e154-4078-89b9-b6e5bb8a4715","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"html","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":78899,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/0984d4c5fd521f3cac95d393.html"},{"id":97795322,"identity":"75f1aae1-c1f9-47e3-af13-aadff4baa42d","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":33680,"visible":true,"origin":"","legend":"\u003cp\u003eSystem architecture diagram\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/31213a3e1080cc2cd429ae6d.png"},{"id":97795302,"identity":"d540fd73-881f-49e5-8593-a221fdbf21d5","added_by":"auto","created_at":"2025-12-09 12:38:44","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":127855,"visible":true,"origin":"","legend":"\u003cp\u003eCNN architecture\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/0d2ef867139677b175ff01c6.png"},{"id":97897460,"identity":"fb9bcdc8-1d87-49e7-81ba-82be4bd9b025","added_by":"auto","created_at":"2025-12-10 15:37:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":888697,"visible":true,"origin":"","legend":"\u003cp\u003ePartial training set images\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/5f7725879d74b62db447c251.png"},{"id":97795348,"identity":"44c174c9-eeab-434b-ae6e-8b4d85cb06fd","added_by":"auto","created_at":"2025-12-09 12:38:46","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":30163,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of model training and validation\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/d02e29e2952244ed9da879a8.png"},{"id":97795371,"identity":"93851237-3456-4f79-a8e8-60e75185cea2","added_by":"auto","created_at":"2025-12-09 12:38:48","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":50076,"visible":true,"origin":"","legend":"\u003cp\u003eAppearance of the integrated terminal device\u003c/p\u003e","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/8596db6f9c6eff63c2928a14.jpg"},{"id":97896076,"identity":"c9ec0aec-2fe2-4cd3-9f49-2343a61fafcf","added_by":"auto","created_at":"2025-12-10 15:35:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":397552,"visible":true,"origin":"","legend":"\u003cp\u003eSimulating a customer placing a product on the weighing sensor\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/ee4a67770e92820ebfda5d0e.png"},{"id":97795354,"identity":"d937df8b-2ef2-44e6-be2a-4d44af8b9174","added_by":"auto","created_at":"2025-12-09 12:38:47","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":298219,"visible":true,"origin":"","legend":"\u003cp\u003ePrinted receipt with barcode\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/11e5dffb540140d715e76e84.png"},{"id":98797626,"identity":"aad8f52c-7e32-40ec-ac6c-381335dcefdf","added_by":"auto","created_at":"2025-12-22 13:36:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2431030,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/36c79d80-f881-4104-860c-bd207f438c1b.pdf"},{"id":97795305,"identity":"05488415-8564-4362-9e0b-4fed940a7ff6","added_by":"auto","created_at":"2025-12-09 12:38:45","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":30884,"visible":true,"origin":"","legend":"","description":"","filename":"Authors.docx","url":"https://assets-eu.researchsquare.com/files/rs-8297422/v1/b78c9bb6b1ff4f83e48bb682.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Centralized Training and Decentralized Invocation: An Innovative Intelligent Recognition and Pricing System for Retail Products","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eIn the retail sector, conventional product recognition and pricing methods often depend on manual entry of product codes or selection from image libraries. These approaches are inefficient, susceptible to errors, and can lead to increased customer wait times and higher operational costs for supermarkets. Furthermore, unmanned stores and self-checkout systems face limitations in recognizing pre-packaged products, particularly in identifying spoilage that is not visible from outside the package. Therefore, developing an intelligent system for accurate and efficient retail product recognition is of paramount importance.\u003c/p\u003e\u003cp\u003ePrior research has explored various techniques. Google and Amazon's recognition services have been utilized for object detection, but these typically do not integrate weighing functionality [1,2]. Ni Yunfeng developed an image acquisition system using cameras and light sources, employing MATLAB to differentiate apples and pears based on color features [3]. Tao Weihua created image acquisition software on a Cortex A8 board to extract color and texture features for fruit and vegetable classification [4]. Yao Xinguang designed a fruit recognition and weighing system using an OPENMV module, identifying fruits through hierarchical scanning of color, size, and similarity [5]. Wang Hui enhanced DarkNet-53 model with YOLOv3 algorithm to improve fruit recognition accuracy [6]. Zou Bin optimized GoogleNet and ResNet-50 structures to develop a fruit recognition model that increased accuracy and reduced processing time [7]. Jin Mei designed an FPGA-based system to meet real-time recognition requirements [8]. Wang Yunfei developed an intelligent electronic scale with an image processing module to identify fruit and vegetable codes for pricing [9\u0026ndash;11]. Zhang Lulu designed a self-service recognition system, replacing the YOLOv5 backbone with a lighter network to boost recognition speed [12]. Chen Dengfeng used pressure sensors and faster R-CNN algorithm, combining image recognition with weight data to generate settlement barcodes for automated checkout [13]. Convolutional Neural Network (CNN)-based intelligent fruits classification utilizing the bilinear pooling with heterogeneous streams is proposed [14].\u003c/p\u003e\u003cp\u003eA common limitation in prior research is that models are typically trained and invoked locally on specific devices. This approach hinders sharing and deployment across different retail terminals, thereby constraining the scope and pace of intelligent upgrades in the industry.\u003c/p\u003e\u003cp\u003eThis paper proposes a paradigm shift: leveraging high-performance cloud computing resources to centrally train deep learning models on large datasets of product images. This process generates models with high recognition precision and recall. In practice, retail terminals can then easily invoke these pre-trained models from the cloud for product recognition, eliminating the need for deploying and maintaining extensive training infrastructure at each terminal. This cloud-centric strategy effectively removes the barriers to the broad and rapid intelligent upgrading of retail systems.\u003c/p\u003e"},{"header":"2 Method","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003e2.1 System architecture\u003c/h2\u003e\n\u003cp\u003eThe proposed system, based on the principle of centralized training and decentralized invocation of deep learning models, integrates computer vision, deep learning, serial communication, thermal printing, and weighing technologies. The overall architecture, depicted in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, consists of the following layers:\u003c/p\u003e\n\u003cp\u003eHardware layer: This includes a pressure sensor (connected to a microcontroller via an HX711AD module), a high-definition camera, a thermal printer, and a computer acting as the control center. The pressure sensor measures weight and transmits data to the computer via serial communication. The camera captures images of products for recognition. The thermal printer generates receipts containing product information.\u003c/p\u003e\n\u003cp\u003eData transmission layer: This layer facilitates data exchange between hardware components through serial ports (USB or RS232). Weight data from the sensor is sent to the microcontroller, which relays it to the computer. The computer encodes images captured by the camera and uploads them to the cloud, subsequently receiving recognition results back from the cloud.\u003c/p\u003e\n\u003cp\u003eCloud-based deep learning model: A Convolutional Neural Network (CNN) model is developed and trained on a cloud AI platform (e.g., Baidu AI Studio) to recognize product types. The input is the product image uploaded from the retail terminal, and the output is the product classification.\u003c/p\u003e\n\u003cp\u003eSoftware layer: An integrated control system, developed in C#, manages camera capture, cloud model invocation, weight data reception, amount calculation, barcode generation, and thermal printer control. This layer also handles serial communication with the microcontroller and pressure sensor.\u003c/p\u003e\n\u003cp\u003eUser interaction layer: When a customer places fruits or vegetables on the electronic scale, the system automatically executes recognition, weighing, pricing, and prints a receipt. This fully automated process significantly enhances shopping efficiency and customer experience.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n\u003ch2\u003e2.2 Technical route\u003c/h2\u003e\n\u003cp\u003eThe technical workflow to achieve product recognition, weighing, pricing, and receipt printing encompasses the following steps:\u003c/p\u003e\n\u003cp\u003eModel construction and training: A CNN model is constructed and trained on a cloud-based deep learning platform using a prepared training set of images. Model parameters are optimized during this phase.\u003c/p\u003e\n\u003cp\u003eSoftware development: An integrated control system is developed in C# to enable communication with hardware devices, invocation of the cloud model, data processing, and printing functions.\u003c/p\u003e\n\u003cp\u003eHardware integration: Electronic barcode scales, cameras, and thermal printers are connected to the computer via USB or serial ports. Necessary configurations and debugging are performed to ensure seamless integration and communication.\u003c/p\u003e\n\u003cp\u003eSystem integration and testing: Comprehensive joint debugging of hardware and software is conducted to ensure all components work harmoniously to complete the required tasks. The system's feasibility, precision, and recall are rigorously verified.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n\u003ch2\u003e2.3 Deep learning model\u003c/h2\u003e\n\u003cdiv id=\"Sec6\" class=\"Section3\"\u003e\n\u003ch2\u003e2.3.1 Model establishment on cloud platform\u003c/h2\u003e\n\u003cp\u003eTo achieve efficient recognition, a CNN model was established and trained on the Baidu AI Open Platform. High-performance server resources, including Tesla V100 16GB GPUs, 12-core CPUs, and 56GB of RAM, were leveraged to ensure training efficiency.\u003c/p\u003e\n\u003cp\u003eThe architecture of the constructed CNN is illustrated in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. The network comprises an input layer, six convolutional layers, a fully connected layer, and a dropout layer. To enhance generalization capability, max-pooling layers are inserted between the last convolutional layer and the fully connected layer. The hidden layers employ the Maxout activation function. The input layer starts with 64 channels, and this number is doubled at each subsequent layer until reaching 512. A flatten function then converts the data into a one-dimensional vector. This vector passes through a 1\u0026times;1\u0026times;512 fully connected layer. To mitigate overfitting, a dropout layer is applied. The final classification result is output by a softmax function.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\n\u003ch2\u003e2.3.2 Dataset preparation\u003c/h2\u003e\n\u003cp\u003eTo ensure model efficacy and robustness, a dataset of 30 types of fruits and vegetables was curated. Images were captured using cameras mounted on the barcode scales, covering single or multiple items, various stacking configurations, and products inside or outside shopping bags. For each type, 200 images were collected. These were randomly split into 120 for training, 40 for validation, and 40 for testing, as detailed in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. Sample images from the training set are shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eComposition of training, validation, and test sets\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eQuantity\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eRequires Tagging?\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eRequires Internet?\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eType\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e30\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e/\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e/\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTraining\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e120 per type\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eYes\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eNo\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eValidation\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e40 per type\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eNo\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eNo\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTest\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e40 per type\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eNo\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eYes\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eTotal\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e6000\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e/\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e/\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\n\u003ch2\u003e2.3.3 Model training and validation\u003c/h2\u003e\n\u003cp\u003eTraining and validation are critical for ensuring the model's precision and recall, forming the foundation for reliable product recognition and pricing. The process is outlined in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eDuring the training phase, the tagged training set (3600 images across 30 categories) was fed into the CNN model. The training involves iterative forward propagation to generate predictions and backpropagation with a gradient descent algorithm to update model parameters, minimizing the error between predictions and actual labels.\u003c/p\u003e\n\u003cp\u003eData augmentation was applied using the Baidu EasyDL platform's default configuration. Hyperparameters were selected via the platform's automatic search feature, and the number of training epochs was set according to its automatic configuration.\u003c/p\u003e\n\u003cp\u003eDuring the validation phase, the untagged validation set (1200 images) was fed into the trained model to evaluate its performance using precision, recall, and the F1-Score, defined as:\u003c/p\u003e\n\u003cp\u003eP\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\text{r}\\text{e}\\text{c}\\text{i}\\text{s}\\text{i}\\text{o}\\text{n}=\\frac{\\text{T}\\text{P}}{\\text{T}\\text{P}+\\text{F}\\text{P}}\\)\u003c/span\u003e\u003c/span\u003e (1)\u003c/p\u003e\n\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ1\" class=\"mathdisplay\"\u003e$$\\:\\text{R}\\text{e}\\text{c}\\text{a}\\text{l}\\text{l}=\\frac{\\text{T}\\text{P}}{\\text{T}\\text{P}+\\text{F}\\text{N}}$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\n\u003cdiv id=\"FileID_Equ2\" class=\"mathdisplay\"\u003e$$\\:{\\text{F}}_{1}-\\text{S}\\text{c}\\text{o}\\text{r}\\text{e}\\:=\\frac{2\\times\\:\\text{P}\\text{r}\\text{e}\\text{c}\\text{i}\\text{s}\\text{i}\\text{o}\\text{n}\\times\\:\\text{R}\\text{e}\\text{c}\\text{a}\\text{l}\\text{l}}{\\text{P}\\text{r}\\text{e}\\text{c}\\text{i}\\text{s}\\text{i}\\text{o}\\text{n}+\\text{R}\\text{e}\\text{c}\\text{a}\\text{l}\\text{l}}$$\u003c/div\u003e\n\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003eWhere:\u003c/p\u003e\n\u003cp\u003eTP (True Positive) is the number of positive samples correctly identified. FP (False Positive) is the number of negative samples incorrectly identified as positive. FN (False Negative) is the number of positive samples incorrectly identified as negative.\u003c/p\u003e\n\u003cp\u003eA validation F1-Score threshold of \u0026ge;\u0026thinsp;99.5% was set. If the model failed to meet this threshold, it underwent further training with hyperparameter adjustments until the target was achieved.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n\u003ch2\u003e2.4 System integration\u003c/h2\u003e\n\u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\n\u003ch2\u003e2.4.1 Weight data acquisition\u003c/h2\u003e\n\u003cp\u003eThe pressure sensor in the electronic scale uses an HX711AD module to convert analog weight signals into digital data. This data is transmitted to the computer via serial communication. A C# program was developed to receive and decode the weight data from the serial port.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section3\"\u003e\n\u003ch2\u003e2.4.2 Image capture\u003c/h2\u003e\n\u003cp\u003eA camera is positioned above the electronic scale. When the load sensor detects an object and transmits weight data to the computer, it simultaneously triggers the camera to capture an image. This image is then sent to the computer for encoding before being submitted to the cloud model for recognition.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section3\"\u003e\n\u003ch2\u003e2.4.3 Integration with cloud model\u003c/h2\u003e\n\u003cp\u003eUpon receiving an image from the camera, the C# program encodes it into Base64 format. It then invokes the cloud-based deep learning model via an API call over the internet. The cloud model processes the image and returns the recognition result to the retail terminal computer.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section3\"\u003e\n\u003ch2\u003e2.4.4 Amount calculation and barcode generation\u003c/h2\u003e\n\u003cp\u003eAfter obtaining the recognition result and weight data, the C# program retrieves the corresponding unit price from a preset database and calculates the total cost. An EAN-128 barcode is generated, encoding product ID, category name, price, and weight using a standard encoding algorithm.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section3\"\u003e\n\u003ch2\u003e2.4.5 receipt printing\u003c/h2\u003e\n\u003cp\u003eThe C# program sends the product details (type name, unit price, weight, total cost) and the generated barcode to the thermal printer. Control code ensures the barcode and text are printed clearly and without errors.\u003c/p\u003e\n\u003cp\u003eThrough careful hardware selection and software design, the electronic scale, camera, and thermal printer are seamlessly integrated into a single terminal device, as shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n\u003ch2\u003e2.5 Testing experiments\u003c/h2\u003e\n\u003cp\u003eDuring the testing phase, the entire system was evaluated using the 1200-image test set. The tests simulated real-world usage, requiring internet connectivity for the terminal device to communicate with the cloud model. The main steps were:\u003c/p\u003e\n\u003c/div\u003e\n\u003cp\u003e1) Simulating a customer placing a fruit or vegetable on the weighing sensor (Fig. 6);\u003c/p\u003e\n\u003cp\u003e2) Receiving weight data from the scale via serial communication;\u003c/p\u003e\n\u003cp\u003e3) Triggering the camera to capture an image of the item (one from the test set);\u003c/p\u003e\n\u003cp\u003e4) Uploading the encoded image to the cloud for recognition by the deep learning model;\u003c/p\u003e\n\u003cp\u003e5) Receiving the product type information from the cloud;\u003c/p\u003e\n\u003cp\u003e6) Retrieving the unit price based on the recognized product code;\u003c/p\u003e\n\u003cp\u003e7) Calculating the total amount based on unit price and weight;\u003c/p\u003e\n\u003cp\u003e8) Generating an electronic barcode;\u003c/p\u003e\n\u003cp\u003e9) Printing a receipt with the barcode (Fig. 7);\u003c/p\u003e\n\u003cp\u003e10) Removing the product from the sensor to prepare for the next customer.\u003c/p\u003e\n\u003cp\u003eThis process is fully automated, requiring manual intervention only for placing and removing product. A sample printed receipt is shown in Fig. 7.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"3 Results","content":"\u003cp\u003eA comparison between this research and prior work [1,2] is presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, highlighting key differences in model training location, invocation location, and scalability.\u003c/p\u003e\u003cp\u003eAfter 1200 tests on 30 different types of fruits and vegetables, where the retail terminal invoked the pre-trained cloud model, the system achieved high recognition precision (95.1%), recall (95.0%), and F1-Score (96.2%). Detailed results for each type are listed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. The heatmap of the confusion matrix is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e. The serial numbers on the axes of Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e correspond to the serial numbers in the first column of Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eMost categories achieved perfect scores. However, cherries and grapes showed a slight probability of being misclassified for one another. A similar confusion was observed between apples and dangshan pears, and between Hami melons and watermelons.\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\u003eComparison with prior research\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=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eModel training location\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eModel invocation location\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eReadily deployable to multiple terminals?\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePrior research\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eLocal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eLocal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNo\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eThis research\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCloud\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eDecentralized terminal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eYes\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\u003eThe results demonstrate the feasibility of the proposed architecture based on centralized cloud training and decentralized invocation for product recognition. Compared to previous studies where both training and recognition were conducted locally on retail terminals [12,13], this approach significantly lowers the barrier to deploying intelligent recognition systems. Electronic barcode scales in various settings\u0026mdash;supermarkets or agricultural markets\u0026mdash;can now leverage advanced AI recognition as long as a basic internet connection is available.\u003c/p\u003e\u003cp\u003eNotably, the system exhibits excellent scalability. Introducing a new product type only requires training the model on this specific type. Once validated, the updated model can be deployed to all terminals, immediately making the new product recognizable across the entire network.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eGrape and cherry are both small in round shape and similar in color, which makes it difficult to completely distinguish between them. Additionally, apple and dangshan pear share a medium-sized, round form with comparable colors and textures, while watermelon and hami melon are both large and spherical, featuring green outer skins. This imposes a constraint on improvement of precision and recall.\u003c/p\u003e\u003cp\u003eThe primary challenges to improving precision and recall stem from visual similarities between certain fruits. Grapes and cherries are both small and round with similar colors. Apples and dangshan pears share a medium-sized round form with comparable colors and textures. Watermelons and hami melons are both large and spherical with green rinds. These inherent visual similarities present the main constraint on further precision and recall improvements.\u003c/p\u003e\u003cp\u003eIn comparison with specific application scenarios merely focusing on recognition [15,16] or classification [17,18], this research realizes the integration of classification and weighing, thereby achieving a more universal scope of application. The intelligent recognition and pricing system, with its cloud-centric design, is readily deployable across multiple retail terminals. It is particularly suitable for chain supermarkets, where unified standards for product types, barcode scales, and environmental lighting can be enforced across locations. By consistently uploading images to a centrally maintained and improved cloud model, high levels of precision and recall can be sustained and enhanced over time.\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\u003eRecognition results on the test set\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=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" 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\u003eNo.\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eType\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTest frequency\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAvg. weight(kg)\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\u003eF1-Score\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\u003eJujube\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9.713\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003ePitaya\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e9.268\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003eCherry\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.975\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e95.0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e96.2%\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\u003ePineapple\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.741\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003eOnion\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.635\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003eKefuli\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e7.515\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003eTomatoes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e7.400\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003ePeach\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6.999\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e97.5%\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\u003ePumpkin\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6.782\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePotato\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6.131\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eJackfruit\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.236\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eKnight Orange\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.177\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDurian\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.915\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eWater-melon\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.144\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eDangshan Pear\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.604\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.6%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e98.8%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGrape\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.076\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e95.1%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e96.3%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eApple\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.972\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.4%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e95.0%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e96.2%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePersim-mon\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.964\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGreen pepper\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.566\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eHami melon\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.457\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e97.5%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMango\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.432\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ePapaya\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.138\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBanana\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.097\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eKiwifruit\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2.068\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStraw-berry\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.979\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCoconut\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.874\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFig\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.649\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eGrapefruit\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1.560\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAvocado\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.852\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBlueberry\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.476\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e100%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e100%\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\u003eThis study successfully demonstrates the principle of centralized training and decentralized invocation for fruit and vegetable recognition. However, it also highlights areas for future investigation. Potential issues related to network latency and data security were not covered in this work. Network delays could impact recognition speed during peak hours, potentially reducing system responsiveness. Furthermore, uploading product images to the cloud necessitates robust data privacy and security measures to protect potentially sensitive information. Investigating methods to reduce latency and enhance data security will be key directions for our future research.\u003c/p\u003e"},{"header":"5 Conclusion","content":"\u003cp\u003eThis research has successfully implemented a centralized training and decentralized invocation framework for fruit and vegetable recognition by establishing a deep learning model on a cloud platform. At the decentralized retail terminals, functionalities for image encoding, uploading, receiving recognition results, and printing settlement barcodes were developed and integrated. The deployment of this system across multiple terminals represents an innovative integration of artificial intelligence with the retail industry, paving the way for the widespread and scalable application of artificial intelligence in this domain.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eYunfei Wang wrote the main manuscript text and Yaping Li prepared figures 1-8. All authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\u003cp\u003eThis research is supported by science and technology project of Henan Province, China (No.162102210320).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eFu Yuesheng, Song Jian, Xie Fuxiang.: Circular Fruit and Vegetable Classification Based on Optimized GoogLeNet, IEEE Access. 9, 113599\u0026ndash;113611(2021)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSharma, V.: Object Detection and Recogni-tion using Amazon Rekognition with Boto3, 6th international conference on trends in electronics and informatics, Tirunelveli, India.727\u0026ndash;732 (2022)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNi Yunfeng, Ye Jian, Fan Jiaojiao.: Fruit sorting system based on image Recognition, Jiangsu Agricultural Sciences. 49 (1), 170\u0026ndash;176 (2021)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTao Weihua, Zhao Li, Xi Ji.: Fruits and vegetables recognition based on color and texture features, Transactions of Chinese Society of Agricultural Engineering. 30(16), 305\u0026ndash;311 (2014)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYao Xinguang, Su Xiulan, Li Qinyu.: Design of Intelligent Fruit Recognition and Weighing System, Electronics World. 43(1), 174\u0026ndash;175(2021)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Hui, Zhang Fan, Liu Xiaofeng.: Fruit image recognition based on DarkNet-53 and YOLOv3, Journal of Northeast Normal Uiversity (Natural Science Edition). 52(4), 60\u0026ndash;65(2020)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZou Bin.: Vegetable-Fruit Recognition and Price Estimating Using Neural Networks, M.A. dissertation, Dept. School of Electronics and Information, Harbin Institute of Technology, Haerbin, Hei Longjiang, China (2023)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJin Mei, Zeng Xin, Zhang Liguo.: A fruit identification system design based on field programmable gate array, Chinese High Technology. 34(6), 616\u0026ndash;623(2024)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Yunfei.: An electronic barcode scale intelligent recognition of bulk commodity item numbers, China patent 2019211701592 (2019)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Yunfei.: A smart membership fingerprint electronic scale for identifying products, China patent 2019216994458 (2019)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang Yunfei.: A smart membership card electronic scale for identifying products, China patent 2019215705817 (2019)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Lulu.: Design of an automatic recognition system for fruits and vegetables based on machine vision, Qingdao University, M.A. dissertation, Dept. School of automation, Qingdao University, Qingdao, Shan Dong, China (2023)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Dengfeng, Zhou Yao, Duan You.: Intelligent fruit and vegetable settlement system based on computer vision, Informatization Research. 45(2) 65\u0026ndash;70 (2019)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePrakash, Achanta Jyothi, Prakasam, P.: An intelligent fruits classification in precision agriculture using bilinear pooling convolutional neural networks, The Visual Computer. 39(5) 1\u0026ndash;17(2022)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuixian Lin, Haidong Deng, Hong Du.: Low visibility underwater biological target detection based on the improved YOLOV5s, The Visual Computer. 41(12) 1\u0026ndash;14(2025)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAnam Nazir, Muhammad Nadeem Cheema, Bin Sheng.: OFF-eNET: An Optimally Fused Fully End-to-End Network for Automatic Dense Volumetric 3D Intracranial Blood Vessels Segmentation, IEEE Transactions on Image Processing. 29 7192\u0026ndash;7202 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang Wen, Ying Zeng, Shuang Liu.: Dsf-net: a dual-stream fusion network integrating structural and detailed features for fundus-based diabetic retinopathy classification, The Visual Computer. 41(15) 1\u0026ndash;13(2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMasood Anum, Sheng Bin, Yang Po.: Automated Decision Support System for Lung Cancer Detection and Classification via Enhanced RFCN with Multilayer Fusion RPN, IEEE Transactions on Industrial Informatics. 46(12) 7791\u0026ndash;7801(2020).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Integration, Cloud, Deep learning, Recognition, Weight","lastPublishedDoi":"10.21203/rs.3.rs-8297422/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8297422/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn the retail industry, efficient product recognition and pricing systems are crucial for enhancing customer experience and operational efficiency. Traditional methods, which often rely on manual input or image selection, are prone to inefficiency and errors. This paper introduces an innovative intelligent system that integrates cameras, electronic scales, and thermal printers with cloud-based deep learning technologies. The core of this approach lies in centralized training of a convolutional neural network (CNN) model on a high-performance cloud platform, enabling decentralized invocation of the recognition results at retail terminals. By uploading product images to the cloud, the system achieves rapid and accurate recognition, with precision, recall, and F1-Score exceeding 95.1%, 95.0%, and 96.2%, respectively. We demonstrate that this architecture not only optimizes the shopping process but also facilitates the large-scale application of artificial intelligence in retail environments. The system's feasibility and efficiency have been validated through extensive testing on 30 types of fruits and vegetables, indicating its strong potential for widespread adoption in supermarkets and agricultural markets.\u003c/p\u003e","manuscriptTitle":"Centralized Training and Decentralized Invocation: An Innovative Intelligent Recognition and Pricing System for Retail Products","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-09 12:38:15","doi":"10.21203/rs.3.rs-8297422/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9b97afae-72ab-406b-aed9-32046a493c84","owner":[],"postedDate":"December 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-12-10T22:14:56+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-09 12:38:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8297422","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8297422","identity":"rs-8297422","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}