E-commerce commodity picture label selection auxiliary method based on artificial intelligence image classification
Through the image classification method based on artificial intelligence, the problems of high labor costs, low efficiency and poor consistency in the process of product image labeling on e-commerce platforms are solved, and efficient and accurate classification of e-commerce product images is achieved, reducing the cost of manual labeling and time.
Patent Information
- Application Number
- CN202411415061.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-11
- Publication Date
- 2025-02-07
AI Technical Summary
The prior art has problems of high labor costs, low efficiency and poor consistency in the process of product image labeling on e-commerce platforms, which makes the image labeling time-consuming and affects the training effect of the classification model.
Using an image classification method based on artificial intelligence, we use select feature data of e-commerce product images, generate data sets, and pre-train and collaborative training of preset network models to realize automatic classification of unlabeled images.
It reduces the cost and time of manual labeling, improves classification accuracy and consistency, and realizes efficient and accurate classification of e-commerce product pictures.
Smart Images
Figure CN119399517A_ABST
Abstract
Description
Background Art
[0002] E-commerce brings a huge amount of pictures, and the speed of new product iteration is very fast. When products are put on the e-commerce platform, each product needs to prepare multiple pictures to complete the production of the product details page. Business personnel need to spend a lot of time to find the correct pictures. Therefore, in order to speed up the above work, different brands need to classify the product pictures and model pictures taken, and label them to facilitate the design, planning, product and other ports to screen pictures.
[0003] Currently, companies can complete classification by training labeling models. This usually requires the use of classification models, which is supervised training. Before training, a certain number of images must be prepared for each label as training sets and test sets. When faced with a huge amount of image materials, preparing training images becomes a very labor-intensive and time-consuming task.
[0004] Existing technologies mainly rely on manual annotation to prepare training images, which has the following problems:
[0005] 1) High labor cost: A large amount of manpower is required for image annotation.
[0006] 2) Low efficiency: Manual labeling is slow and cannot meet the processing requirements of large-scale images.
[0007] 3) Poor consistency: The standards of different labelers are inconsistent, which affects the training effect of the classification model. Summary of the invention
[0008] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and specifically provide the following technical solutions:
[0009] 1) In the first aspect, the present invention provides an auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification, and the specific technical solution is as follows:
[0010] Select at least one e-commerce product image from each category and label it, and obtain feature data for each selected e-commerce product image;
[0011] Generate a data set including multiple samples, where the label and feature data of any selected e-commerce product image constitute a sample;
[0012] Using the data set, pre-training the first preset network model and the second preset network model respectively to obtain the pre-trained first preset network model and the pre-trained second preset network model;
[0013] Co-training the pre-trained first preset network model and the pre-trained second preset network model to obtain a trained first preset network model and a trained second preset network model;
[0014] Using the trained first preset network model or the trained second preset network model to classify all unlabeled e-commerce product images to obtain preliminary classification results;
[0015] Determine whether the preliminary classification result meets the first preset requirement and obtain a determination result;
[0016] When the judgment result is yes, the classification of all e-commerce product images is completed.
[0017] Based on the above scheme, the auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification of the present invention can also be improved as follows.
[0018] Further, the pre-trained first preset network model and the pre-trained second preset network model are collaboratively trained to obtain the trained first preset network model and the trained second preset network model, including:
[0019] Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training; based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training; the second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training; based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
[0020] Furthermore, it also includes:
[0021] When the judgment result is no, based on the preliminary classification results, at least one e-commerce product image of each category is selected from all unlabeled e-commerce product images and labeled, and the feature data of each selected e-commerce product image is obtained, and multiple new samples are generated and added to the data set, and the model is re-trained until the judgment result is yes.
[0022] Furthermore, the process of obtaining the feature data of each selected e-commerce product image includes:
[0023] The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
[0024] 2) In the second aspect, the present invention also provides an e-commerce product image label selection auxiliary system based on artificial intelligence image classification, and the specific technical solution is as follows:
[0025] It includes a labeling and feature extraction module, a data set generation module, a model pre-training module, a model collaborative training module, a classification module, a judgment module and a determination module;
[0026] The labeling and feature extraction module is used to: select at least one e-commerce product image of each category and label it, and obtain feature data of each selected e-commerce product image;
[0027] The data set generation module is used to: generate a data set including multiple samples, wherein the label and feature data of any selected e-commerce product image constitute a sample;
[0028] The model pre-training module is used to: use the data set to pre-train the first preset network model and the second preset network model respectively, to obtain the pre-trained first preset network model and the pre-trained second preset network model;
[0029] The model collaborative training module is used to: collaboratively train the pre-trained first preset network model and the pre-trained second preset network model to obtain the trained first preset network model and the trained second preset network model;
[0030] The classification module is used to: classify all unlabeled e-commerce product images using the trained first preset network model or the trained second preset network model to obtain preliminary classification results;
[0031] The judgment module is used to: judge whether the preliminary classification result meets the first preset requirement and obtain a judgment result;
[0032] The determination module is used to: when the judgment result is yes, complete the classification of all e-commerce product images.
[0033] Based on the above scheme, the e-commerce product image label selection auxiliary system based on artificial intelligence image classification of the present invention can also be improved as follows.
[0034] Furthermore, the model collaborative training module is specifically used for:
[0035] Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training; based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training; the second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training; based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
[0036] Furthermore, a repeated calling module is included, and the repeated calling module is used to:
[0037] When the judgment result is no, the labeling and feature extraction module is called according to the preliminary classification result, and at least one e-commerce product image of each category is selected and labeled from all unlabeled e-commerce product images, and the feature data of each selected e-commerce product image is obtained, and multiple new samples are generated and added to the data set, and the model pre-training module, model collaborative training module, classification module, judgment module and determination module are re-called to re-train the model until the judgment result is yes.
[0038] Furthermore, the annotation and feature extraction module is also specifically used for:
[0039] The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
[0040] 3) In a third aspect, the present invention further provides an electronic device, comprising a processor, the processor being coupled to a memory, the memory storing at least one computer program, and the at least one computer program being loaded and executed by the processor, so that the electronic device can implement any of the above-mentioned e-commerce product image label selection auxiliary methods based on artificial intelligence image classification.
[0041] 4) In a fourth aspect, the present invention further provides a computer-readable storage medium, in which at least one computer program is stored, and at least one computer program is loaded and executed by a processor so that a computer implements any of the above-mentioned e-commerce product image label selection auxiliary methods based on artificial intelligence image classification.
[0042] The beneficial effects of the present invention are as follows:
[0043] The business personnel responsible for the background image data processing can first select a very small number of e-commerce product images of each category and mark them, and combine them with corresponding feature data to form a data set, and use the data set to pre-train the first preset network model and the second preset network model respectively, and then collaboratively train the pre-trained first preset network model and the pre-trained second preset network model, thereby improving the classification accuracy of the trained first preset network model and the trained second preset network model, and use the trained first preset network model or the trained second preset network model to classify all unlabeled e-commerce product images to obtain preliminary classification results. When the preliminary classification results do not meet the preset requirements, re-select and generate new samples based on the unlabeled e-commerce product images of each category, and then perform model training again, and then use the trained first preset network model or the trained second preset network model obtained again to classify all unlabeled e-commerce product images, and repeat the process in subsequent processing. At this time, the classification accuracy will gradually increase until the classification result meets the preset requirements. The present invention realizes accurate classification of e-commerce product images in a semi-supervised learning manner, greatly reducing labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:
[0045] Figure 1 A flowchart of an auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification according to an embodiment of the present invention;
[0046] Figure 2 This is a structural schematic diagram of an e-commerce product image label selection auxiliary system based on artificial intelligence image classification according to an embodiment of the present invention;
[0047] Figure 3 The figure is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0048] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.
[0049] like Figure 1 As shown, a flowchart of an auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification according to an embodiment of the present invention includes the following steps:
[0050] S1. Select at least one e-commerce product image from each category and label it, and obtain feature data of each selected e-commerce product image. The feature data is a multi-dimensional vector feature, which can be obtained by extracting features from the e-commerce product images using a trained ViT model.
[0051] Among them, the e-commerce product pictures can be the front picture, back picture, bottom picture and top picture of a certain product. The categories are divided into: front, back, bottom and top. The marked labels can be the category names. The categories of e-commerce product pictures and the marked labels can also be set according to actual conditions.
[0052] S2. Generate a data set including multiple samples, wherein the label and feature data of any selected e-commerce product image constitute a sample;
[0053] S3. Pre-train the first preset network model and the second preset network model respectively using the data set to obtain the pre-trained first preset network model and the pre-trained second preset network model;
[0054] The first preset network model and the second preset network model can be neural networks, or other deep learning models can be selected according to actual conditions. The first preset network model and the second preset network model have the same network structure.
[0055] Among them, the inputs of the first preset network model and the second preset network model are both feature data of e-commerce product images, and the outputs of the first preset network model and the second preset network model are both predicted labels. There is a corresponding relationship between labels and categories. Therefore, the predicted labels can realize the classification of e-commerce product images.
[0056] S4, collaboratively training the pre-trained first preset network model and the pre-trained second preset network model to obtain a trained first preset network model and a trained second preset network model;
[0057] S5. Classify all unlabeled e-commerce product images using the trained first preset network model or the trained second preset network model to obtain preliminary classification results;
[0058] S6. Determine whether the preliminary classification result meets the first preset requirement and obtain a determination result;
[0059] S7. When the judgment result is yes, complete the classification of all e-commerce product images, specifically:
[0060] According to the preliminary classification results, multiple e-commerce product images are selected from each category, and it is manually judged whether they are correctly classified. When the classification accuracy of all the sampled e-commerce product images reaches the preset classification accuracy threshold, the preliminary classification results are judged to meet the first preset requirement. At this time, the judgment result is yes. If the classification accuracy of all the sampled e-commerce product images does not reach the preset classification accuracy threshold, the preliminary classification results are judged to not meet the first preset requirement. At this time, the judgment result is no, wherein the preset classification accuracy threshold can be set according to actual conditions.
[0061] Among them, completing the classification of all e-commerce product images means: completing the classification of some e-commerce product images through S1, and completing the classification of the remaining e-commerce product images (that is, all unlabeled e-commerce product images except the e-commerce product images labeled in S1) through S5.
[0062] Optionally, in S4, the pre-trained first preset network model and the pre-trained second preset network model are collaboratively trained to obtain the trained first preset network model and the trained second preset network model, including:
[0063] S40. Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training. Based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training. The second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training. Based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
[0064] Among them, the second preset requirement can be set according to actual conditions. For example, the second preset requirement is: when training the first model and the second model, the loss values of the loss functions used are less than the preset loss value threshold, wherein the preset loss value threshold can be set according to actual conditions.
[0065] The first model and the second model are used to generate pseudo labels simultaneously or at different times. Subsequently, the first model after each training and the second model after each training are used to generate pseudo labels simultaneously or at different times.
[0066] Optionally, in the above technical solution, it also includes:
[0067] S8. When the judgment result is no, based on the preliminary classification result, select at least one e-commerce product image of each category from all unlabeled e-commerce product images and label them, obtain feature data of each selected e-commerce product image, generate multiple new samples, add them to the data set, and re-train the model until the judgment result is yes.
[0068] Among them, re-training the model can be implemented in two ways:
[0069] 1) The first implementation method: using the expanded data set, return to execute S3, re-pre-train the first preset network model and the second preset network model, and then execute subsequent steps.
[0070] 2) The second implementation method: using the expanded data set, taking the currently obtained trained first preset network model as the first preset network model, and taking the currently obtained trained second preset network model as the second preset network model, and returning to execute S3.
[0071] Business personnel responsible for background image data processing can select and label at least one e-commerce product image in each category based on the preliminary classification results, and expand the number of samples in the data set. As the number of samples increases, the classification accuracy of the trained first preset network model or the trained second preset network model will be improved.
[0072] On several different data sets (including data sets of multiple e-commerce product images), with the increase in the number of labeled e-commerce product images, the classification accuracy of e-commerce product images has also steadily increased. The initial classification accuracy can reach about 20%. After repeated selection and repeated training for 7 to 10 rounds, the classification accuracy gradually increases to 80% to 90%.
[0073] Optionally, in S1, the process of obtaining feature data of each selected e-commerce product image includes:
[0074] The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
[0075] Among them, the trained ViT model specifically refers to: a pre-trained model (trained model) related to Vision Transformer. Vision Transformer can be a model such as DINOv2_b, DINOv2_h, DINOv2_g or CLIP. The ViT model can batch compress multiple e-commerce product images and obtain feature data of multiple e-commerce product images in batches.
[0076] The beneficial effects of the present invention are as follows:
[0077] 1) The business personnel responsible for the background image data processing can first select a very small number of e-commerce product images of each category and mark them, and combine them with corresponding feature data to form a data set, and use the data set to pre-train the first preset network model and the second preset network model respectively, and then collaboratively train the pre-trained first preset network model and the pre-trained second preset network model, thereby improving the classification accuracy of the trained first preset network model and the trained second preset network model, and use the trained first preset network model or the trained second preset network model to classify all unlabeled e-commerce product images to obtain preliminary classification results. When the preliminary classification results do not meet the preset requirements, re-select and generate new samples based on the unlabeled e-commerce product images of each category, and then perform model training again, and then use the trained first preset network model or the trained second preset network model obtained again to classify all unlabeled e-commerce product images, and repeat the process in subsequent processing. At this time, the classification accuracy will gradually increase until the classification result meets the preset requirements. The present invention realizes accurate classification of e-commerce product images in a semi-supervised learning manner, greatly reducing labor costs.
[0078] 2) In actual application, if the entire data set contains 10,000 e-commerce product images, each image classification can be completed in less than 10 seconds, and manual selection and judgment can be performed, which greatly reduces the overall time cost.
[0079] 3) Strong data integration capabilities, able to process e-commerce product images from different sources, and improve the utilization and efficiency of e-commerce product images.
[0080] 4) Efficient labeling: Automatic classification is performed through the trained first preset network model or the trained second preset network model, greatly improving the classification efficiency.
[0081] 5) High labeling consistency, reducing human errors and improving model training effects.
[0082] 6) Low cost, reducing reliance on manual labeling and reducing cost overhead.
[0083] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given by the present invention. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.
[0084] like Figure 2As shown, an e-commerce product image label selection auxiliary system 200 based on artificial intelligence image classification according to an embodiment of the present invention includes a labeling and feature extraction module 201, a data set generation module 202, a model pre-training module 203, a model collaborative training module 204, a classification module 205, a judgment module 206 and a determination module 207;
[0085] The labeling and feature extraction module 201 is used to: select at least one e-commerce product image of each category and label it, and obtain feature data of each selected e-commerce product image;
[0086] The data set generation module 202 is used to: generate a data set including a plurality of samples, wherein the label and feature data of any selected e-commerce product image constitute a sample;
[0087] The model pre-training module 203 is used to: use the data set to pre-train the first preset network model and the second preset network model respectively, to obtain the pre-trained first preset network model and the pre-trained second preset network model;
[0088] The model collaborative training module 204 is used to: collaboratively train the pre-trained first preset network model and the pre-trained second preset network model to obtain the trained first preset network model and the trained second preset network model;
[0089] The classification module 205 is used to classify all unlabeled e-commerce product images using the trained first preset network model or the trained second preset network model to obtain a preliminary classification result;
[0090] The judgment module 206 is used to: judge whether the preliminary classification result meets the first preset requirement, and obtain a judgment result;
[0091] The determination module 207 is used to: when the judgment result is yes, complete the classification of all e-commerce product images.
[0092] Optionally, in the above technical solution, the model collaborative training module 204 is specifically used for:
[0093] Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training; based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training; the second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training; based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
[0094] Optionally, in the above technical solution, a repeated calling module is further included, and the repeated calling module is used to:
[0095] When the judgment result is no, based on the preliminary classification result, the labeling and feature extraction module 201 is called to select at least one e-commerce product image of each category from all unlabeled e-commerce product images and label them, and obtain the feature data of each selected e-commerce product image, generate multiple new samples, and add them to the data set, and re-call the model pre-training module, model collaborative training module, classification module, judgment module and determination module, and re-train the model until the judgment result is yes.
[0096] Optionally, in the above technical solution, the labeling and feature extraction module 201 is further specifically used for:
[0097] The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
[0098] It should be noted that the beneficial effects of the e-commerce product image label selection auxiliary system 200 based on artificial intelligence image classification provided by the above embodiment are the same as the beneficial effects of the above-mentioned e-commerce product image label selection auxiliary method based on artificial intelligence image classification, which will not be repeated here. In addition, when the system provided by the above embodiment realizes its functions, it only uses the division of the above-mentioned functional modules as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the system can be divided into different functional modules according to actual conditions to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.
[0099] like Figure 3As shown, an electronic device 300 of an embodiment of the present invention includes a processor 320, the processor 320 is coupled to a memory 310, and the memory 310 stores at least one computer program 330, and the at least one computer program 330 is loaded and executed by the processor 320, so that the electronic device 300 implements any of the above-mentioned auxiliary methods for selecting e-commerce product image labels based on artificial intelligence image classification, specifically:
[0100] The electronic device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors 320 (Central Processing Units, CPU) and one or more memories 310, wherein the one or more memories 310 store at least one computer program 330, and the at least one computer program 330 is loaded and executed by the one or more processors 320, so that the electronic device 300 implements any of the e-commerce product image label selection auxiliary methods based on artificial intelligence image classification provided in the above embodiments. Of course, the electronic device 300 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The electronic device 300 may also include other components for realizing device functions, which will not be repeated here. The electronic device may specifically be a computer, etc.
[0101] A computer-readable storage medium according to an embodiment of the present invention stores at least one computer program, and the at least one computer program is loaded and executed by a processor so that a computer implements any of the above-mentioned auxiliary methods for selecting e-commerce product image labels based on artificial intelligence image classification.
[0102] Optionally, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, and the like.
[0103] In an exemplary embodiment, a computer program product or a computer program is also provided, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes any of the above-mentioned methods for assisting e-commerce product image label selection based on artificial intelligence image classification.
[0104] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects and represent the definition of a specific order or sequence. The order of use of similar objects can be interchanged where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than the order shown or described.
[0105] Those skilled in the art know that the present invention can be implemented as a system, method or computer program product. Therefore, the present invention can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" herein. In addition, in some embodiments, the present invention can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.
[0106] Any combination of one or more computer-readable media can be used. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.
[0107] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. An auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification, characterized in that: include: Select at least one e-commerce product image from each category and label it, and obtain feature data for each selected e-commerce product image; Generate a data set including multiple samples, where the label and feature data of any selected e-commerce product image constitute a sample; Using the data set, pre-training the first preset network model and the second preset network model respectively to obtain the pre-trained first preset network model and the pre-trained second preset network model; Co-training the pre-trained first preset network model and the pre-trained second preset network model to obtain a trained first preset network model and a trained second preset network model; Using the trained first preset network model or the trained second preset network model to classify all unlabeled e-commerce product images to obtain preliminary classification results; Determine whether the preliminary classification result meets the first preset requirement, and obtain a determination result; When the judgment result is yes, the classification of all e-commerce product images is completed.
2. According to claim 1, the auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification is characterized in that: The pre-trained first preset network model and the pre-trained second preset network model are collaboratively trained to obtain the trained first preset network model and the trained second preset network model, including: Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training; based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training; the second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training; based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
3. The method for selecting e-commerce product image labels based on artificial intelligence image classification according to claim 1 or 2, characterized in that: Also includes: When the judgment result is no, based on the preliminary classification result, at least one e-commerce product image of each category is selected from all unlabeled e-commerce product images and labeled, and feature data of each selected e-commerce product image is obtained, and multiple new samples are generated and added to the data set, and the model is re-trained until the judgment result is yes.
4. The method for selecting e-commerce product image labels based on artificial intelligence image classification according to claim 1 or 2, characterized in that: The process of obtaining the feature data of each selected e-commerce product image includes: The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
5. An e-commerce product image label selection auxiliary system based on artificial intelligence image classification, characterized in that: It includes a labeling and feature extraction module, a data set generation module, a model pre-training module, a model collaborative training module, a classification module, a judgment module and a determination module; The labeling and feature extraction module is used to: select at least one e-commerce product image of each category and label it, and obtain feature data of each selected e-commerce product image; The data set generation module is used to: generate a data set including a plurality of samples, wherein the label and feature data of any selected e-commerce product image constitute a sample; The model pre-training module is used to: use the data set to pre-train the first preset network model and the second preset network model respectively, to obtain the pre-trained first preset network model and the pre-trained second preset network model; The model collaborative training module is used to: collaboratively train the pre-trained first preset network model and the pre-trained second preset network model to obtain the trained first preset network model and the trained second preset network model; The classification module is used to classify all unlabeled e-commerce product images using the trained first preset network model or the trained second preset network model to obtain a preliminary classification result; The judgment module is used to: judge whether the preliminary classification result meets the first preset requirement, and obtain a judgment result; The determination module is used to: when the judgment result is yes, complete the classification of all e-commerce product pictures.
6. The e-commerce product image label selection auxiliary system based on artificial intelligence image classification according to claim 5 is characterized in that: The model collaborative training module is specifically used for: Based on the pseudo labels generated by the first model, the second model is trained to obtain the second model after the first training; based on the pseudo labels generated by the second model, the first model is trained to obtain the first model after the first training; the second model after the first training is trained using the pseudo labels generated by the first model after the first training to obtain the second model after the second training; based on the pseudo labels generated by the second model after the first training, the first model after the first training is trained until it meets the second preset requirement, wherein the first model is the first preset network model after pre-training, and the second model is the second preset network model after pre-training, or the first model is the second preset network model after pre-training, and the second model is the first preset network model after pre-training.
7. The e-commerce product image label selection auxiliary system based on artificial intelligence image classification according to claim 5 or 6, characterized in that: It also includes a repeated calling module, wherein the repeated calling module is used to: When the judgment result is no, the labeling and feature extraction module is called according to the preliminary classification result, and at least one e-commerce product image of each category is selected and labeled from all unlabeled e-commerce product images, and the feature data of each selected e-commerce product image is obtained, and multiple new samples are generated and added to the data set, and the model pre-training module, model collaborative training module, classification module, judgment module and determination module are re-called, and the model training is re-performed until the judgment result is yes.
8. The e-commerce product image label selection auxiliary system based on artificial intelligence image classification according to claim 5 or 6, characterized in that: The labeling and feature extraction module is also specifically used for: The trained ViT model is used to extract features of each selected e-commerce product image to obtain feature data of each selected e-commerce product image.
9. An electronic device, characterized in that: The electronic device includes a processor, the processor is coupled to a memory, and at least one computer program is stored in the memory. The at least one computer program is loaded and executed by the processor so that the electronic device implements an e-commerce product image label selection auxiliary method based on artificial intelligence image classification as described in any one of claims 1 to 4.
10. A computer-readable storage medium, characterized in that: At least one computer program is stored in the computer-readable storage medium, and the at least one computer program is loaded and executed by the processor so that the computer implements an auxiliary method for selecting e-commerce product image labels based on artificial intelligence image classification as described in any one of claims 1 to 4.