Vector-based platform picture searching and displaying method and device and medium
Through the deep learning model, product features are extracted and partitioned searches are performed, the problem of poor search results in e-commerce platform image search is solved when angle differences are different, efficient and accurate image search and personalized recommendations are achieved, and user experience and purchase conversion rate are improved.
Patent Information
- Application Number
- CN202411934768.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-26
- Publication Date
- 2025-03-25
AI Technical Summary
The existing e-commerce platform image search technology does not have good search results when there are angle differences in product images, and vector-based image search requires a lot of calculations, resulting in slower search speed and can only be displayed in personalized form.
Product features are extracted through deep learning models and represented as high-dimensional feature vectors, pre-classified to obtain product types, partition searches to improve search accuracy, and personalized sorting is performed based on user history.
It significantly improves search accuracy, supports large-scale data retrieval and ensures system stability, realizes personalized recommendations, improves user experience and increases purchase conversion rate.
Smart Images
Figure CN119691257A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of e-commerce platforms, and in particular to a vector-based platform image search and display method, device and medium. Background Art
[0002] When searching for products on e-commerce platforms, users often face the problem of not being able to accurately describe the products they need, resulting in unsatisfactory search results. In addition, some users hope to search for similar products by uploading pictures instead of relying on text descriptions. Image search technology has been widely used on major e-commerce platforms, providing users with a convenient way to search for products.
[0003] With the continuous development of technologies such as artificial intelligence and deep learning, the image search technology of e-commerce platforms is also continuously upgraded. Although the image search technology of e-commerce platforms has made certain progress, in some cases, the accuracy of search results still needs to be improved. When there are differences in angle, light or background of product images, the search effect may be poor; vector-based AI image search requires a lot of calculations, including image feature extraction, vector similarity calculation, etc. These calculation processes have high requirements for computing resources and may cause the search speed to slow down or the system to be unstable. These problems will affect users' enthusiasm for buying and affect sales.
[0004] Through the above analysis, the problems and defects of the prior art are as follows: The existing image search technology may lead to poor search results when there are angle differences in product images. Vector-based image search requires a lot of calculations, which slows down the search speed and can only be displayed based on personalization. Summary of the invention
[0005] The embodiments of the present application provide a vector-based platform image search and display method, device and medium for solving the above-mentioned technical problems.
[0006] In a first aspect, an embodiment of the present application provides a vector-based platform image search and display method, characterized in that the method includes: receiving, through an e-commerce platform interface, products to be searched uploaded by users, wherein there is at least one product to be searched and the product to be searched includes at least one product image or product link; using a deep learning model to extract product features of the product to be searched, and representing the product features as a first feature vector; pre-classifying the first feature vector to obtain product categories of the product to be searched; obtaining an area to be searched according to the product category, and performing partitioned retrieval on the product to be searched through the area to be searched to obtain a target area; retrieving products for sale similar to the product to be searched in the target area, and asking the user whether to display them in a personalized manner.
[0007] In one implementation of the present application, before receiving the searched goods uploaded by the user through the e-commerce platform interface, the method also includes: converting the image and description text of the goods for sale into a high-dimensional vector to obtain a second feature vector; forming data points of the second feature vector in the vector space, each data point corresponding to an image and description text of the goods for sale; constructing a star map structure unit according to the similarity between the data points, and the connection between the data points represents the similarity relationship between the images; in the star map structure unit, selecting a first preset number of data points as the starting point, expanding to the surrounding according to the similarity relationship to obtain a second preset number of data points, and aggregating them together to obtain the product category.
[0008] In one implementation of the present application, an area to be searched is obtained according to the type of goods, and the goods to be searched are partitioned and retrieved through the area to be searched to obtain a target area, specifically including: selecting data points in the middle of the area to be searched as entry points, and gradually expanding to the data points around them along the similarity relationship; using storage and calculation asynchronous retrieval to simultaneously read the image and calculate the distance, and calculate the similarity between the first eigenvector and the second eigenvector; selecting the data point with the largest similarity as the target area.
[0009] In one implementation of the present application, goods for sale similar to the goods to be searched are retrieved in a target area, and the user is asked whether to display them in a personalized manner, specifically including: performing a similarity calculation in the target area to obtain a preset number of first search results; in the case where the user chooses personalized sorting, dynamically adjusting the search results and displaying a second search result; in the case where the user chooses a repeat prompt, asking the user whether to sort them in a personalized manner.
[0010] In one implementation of the present application, items for sale similar to the item to be searched are retrieved in a target area, and the user is asked whether a personalized display is desired, specifically including: obtaining the user's historical search history, purchase history, and collection history, as well as the user's feedback, and using a machine learning algorithm to construct a user portrait; asking the user whether personalized sorting is performed based on the user portrait, and asking whether the prompt is repeated.
[0011] In one implementation of the present application, the method also includes: when the product category is not found, asking the user whether to specify a local area, extracting local features of the local area through an attention mechanism, and obtaining a first feature map; performing channel compression on the product features of the searched product to obtain two second feature maps containing spatial information; splicing the second feature maps, and performing feature extraction and fusion through a convolutional layer to obtain a 2-channel third feature map; converting the third feature map into a spatial attention weight map, and multiplying it element-by-element with the first feature map to relax or change the product features.
[0012] In one implementation of the present application, the method further includes: when a user clicks on a local area, extracting local features of the local area through a small convolution module; and performing batch pre-classification on multiple product images to obtain multiple product categories.
[0013] In one implementation of the present application, the products to be searched uploaded by the user are received through the e-commerce platform interface, which specifically includes: scanning the products to be searched in the ES server and the image library through the image scanning subsystem; writing the image data into a preset directory through the image writing subsystem; downloading the images on a large scale in the preset directory through the image downloading subsystem, performing preprocessing, and generating a request for the image feature extraction task.
[0014] In the second aspect, an embodiment of the present application also provides a vector-based platform image search and display device, characterized in that the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the at least one processor can: receive, through an e-commerce platform interface, products to be searched uploaded by users, wherein there is at least one product to be searched, and the product to be searched includes at least one product image or product link; use a deep learning model to extract product features of the product to be searched, and represent the product features as a first feature vector; pre-classify the first feature vector to obtain the product categories of the product to be searched; obtain an area to be searched according to the product category, and perform partition retrieval on the product to be searched through the area to be searched to obtain a target area; retrieve products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized way.
[0015] In a third aspect, an embodiment of the present application further provides a non-volatile computer storage medium for vector-based platform image search and display, storing computer executable instructions, characterized in that the computer executable instructions are configured to: receive, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein there is at least one product to be searched and the product to be searched includes at least one product image or product link; use a deep learning model to extract product features of the product to be searched, and represent the product features as a first feature vector; pre-classify the first feature vector to obtain product categories of the product to be searched; obtain an area to be searched according to the product category, and perform partition retrieval on the product to be searched through the area to be searched to obtain a target area; retrieve products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized manner.
[0016] The vector-based platform image search and display method, device and medium provided in the embodiments of the present application have the following beneficial effects: 1. Significantly improve search accuracy: By adopting deep learning models and vector processing technology, this application can convert product images into high-dimensional feature vectors and achieve efficient retrieval in vector databases, which not only overcomes the limitations of traditional keyword search in product descriptions, but also greatly improves the accuracy of search results. Users can quickly find the products they need by uploading pictures, whether it is the style, color of clothing or the design style of furniture, they can get accurate matches.
[0017] 2. Support large-scale data retrieval and ensure system stability: With the increasing variety of goods on e-commerce platforms, traditional search methods often seem powerless when faced with massive amounts of data. This application uses vector databases and efficient vector retrieval tools to easily cope with the challenges of large-scale data retrieval. At the same time, by optimizing database scalability and index design, the system's performance is ensured to be stable, and even with a growing product library, it can maintain fast search speeds and accurate recommendation results.
[0018] 3. Personalized recommendations to improve user experience: This application combines the user's historical browsing history and purchase preferences to personalize the search results. The personalized recommendation method not only improves the user's shopping experience, but also increases the user's purchase conversion rate. Users can find products that meet their needs more quickly, and can also discover more products that may be of interest, thereby enhancing the user's stickiness and loyalty to the e-commerce platform.
[0019] 4. Ensure data security and privacy: In terms of data security, this application uses advanced data encryption and access control technologies to ensure the security and privacy of user data and product information. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 A flow chart of a vector-based platform image search and display method provided in an embodiment of the present application; Figure 2 A schematic diagram of the internal structure of a vector-based platform image search and display device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0021] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.
[0022] The embodiments of the present application provide a vector-based platform image search and display method, device and medium to solve the following technical problems: the image search technology in the prior art may lead to poor search results when there are angle differences in product images. The vector-based image search requires a lot of calculations, which slows down the search speed and can only be displayed based on personalization.
[0023] The technical solution proposed in the embodiments of the present application is described in detail below with reference to the accompanying drawings.
[0024] Figure 1 A flow chart of a vector-based platform image search and display method provided in an embodiment of the present application. Figure 1 As shown, a vector-based platform image search and display method provided in an embodiment of the present application specifically includes the following steps: Step 101: Receive the products to be searched uploaded by the user through the e-commerce platform interface.
[0025] First of all, it should be explained that in the embodiment of the present application, there is at least one product to be searched, and the product to be searched includes at least one product image or product link.
[0026] In one embodiment of the present application, before receiving a product to be searched uploaded by a user through an e-commerce platform interface, where the product to be searched includes a product image and a product link, the method further includes: converting the image and description text of the product to be sold into a high-dimensional vector to obtain a second feature vector; forming data points of the second feature vector in a vector space, where each data point corresponds to an image and description text of the product to be sold; constructing a star map structure unit based on the similarity between the data points, where the connection between the data points represents the similarity relationship between the images; in the star map structure unit, selecting a first preset number of data points as the starting point, expanding to the surrounding areas based on the similarity relationship to obtain a second preset number of data points, and aggregating them together to obtain the product category.
[0027] It is understandable that the purpose of converting the images and description texts of the products for sale into high-dimensional vectors is to encode the complex information in the images and texts into a mathematical form that is easy for computers to process.
[0028] In one embodiment of the present application, the products to be searched uploaded by the user are received through the e-commerce platform interface, which specifically includes: scanning the products to be searched in the ES server and the image library through the image scanning subsystem; writing the image data into a preset directory through the image writing subsystem; downloading the images in a preset directory on a large scale through the image downloading subsystem, performing preprocessing, and generating a request for the image feature extraction task.
[0029] Step 102: Use a deep learning model to extract product features of the product to be searched, and represent the product features as a first feature vector.
[0030] In one embodiment, the deep learning model can be a convolutional neural network (CNN), a residual network (ResNet) or other models suitable for image feature extraction. By training these models, they can learn to extract useful features from product images. When extracting features, the image of the product to be searched is input into the deep learning model, and the model outputs a high-dimensional feature vector, namely the first feature vector, which contains the key information of the product image for subsequent similarity matching and product classification.
[0031] Step 103: pre-classify the first feature vector to obtain commodity categories of the commodity to be searched.
[0032] In one embodiment, pre-classification can adopt a classification algorithm based on machine learning, such as support vector machine (SVM), random forest, etc. First, it is necessary to construct a training data set containing multiple types of goods and label each type of goods. Then, use this training data set to train the classifier. During the training process, the classifier will learn how to map the first feature vector to the corresponding type of goods. After the training is completed, the first feature vector of the product to be searched is input into the classifier, and the classifier will output the type of goods to which the product belongs. In this way, the searched goods can be quickly and accurately pre-classified, which facilitates subsequent search and display. In addition, in order to further improve the accuracy of pre-classification, an integrated learning method can also be used to fuse the output results of multiple classifiers to obtain a more reliable judgment of the type of goods.
[0033] Step 104: Obtain a to-be-searched area according to the commodity types, and perform a partition search on the to-be-searched commodities through the to-be-searched area to obtain a target area.
[0034] In one embodiment of the present application, after obtaining the commodity category of the commodity to be searched, the area to be searched is obtained according to the commodity category, and the commodity to be searched is partitioned and retrieved through the area to be searched to obtain the target area.
[0035] Specifically, the data points in the middle of the area to be searched are selected as the entry point, and are gradually expanded to the surrounding data points along the similarity relationship; asynchronous retrieval is used to simultaneously read the image and calculate the distance to calculate the similarity between the first eigenvector and the second eigenvector; the data point with the greatest similarity is selected as the target area.
[0036] In one embodiment, a rough area to be searched can be determined according to the commodity category obtained in step 103. This area may contain all commodity data related to the commodity category, or an optimized search area obtained based on the correlation analysis of historical data and commodity categories. In the area to be searched, one or more data points are selected as the starting point (entry point) of the retrieval; usually, these entry points may be the center point of the area, or representative data points selected according to a certain strategy (such as density-based clustering center). Starting from the entry point, the search range is gradually expanded along the similarity relationship between the data points (these relationships have been determined when the star map structure unit is constructed in step 101) to explore the surrounding data points. In the retrieval process, the storage-computation asynchronous retrieval technology is used, that is, data reading (obtaining the feature vector of the commodity image from the storage system) and distance calculation (calculating the similarity between the first feature vector of the commodity to be searched and the second feature vector in the storage) are performed simultaneously. The asynchronous processing method can significantly improve the retrieval efficiency because it avoids the delay of waiting for data reading when calculating the similarity. For each retrieved data point, the system calculates the similarity between its second feature vector and the first feature vector of the commodity to be searched. This is usually achieved by calculating the cosine similarity, Euclidean distance or other similarity measures between two vectors. After comparing the similarities of all relevant data points, the area where the data points with the greatest similarity are located is selected as the target area. This target area contains the product data that is most similar to the product to be searched.
[0037] Step 105: Retrieve the products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized way.
[0038] In one embodiment of the present application, after the target area is obtained, the goods for sale similar to the goods to be searched are retrieved in the target area, and the user is asked whether to display them in a personalized manner.
[0039] Specifically, similarity calculation is performed in the target area to obtain a preset number of first search results; when the user selects personalized sorting, the search results are dynamically adjusted and the second search results are displayed; when the user selects repeat prompts, the user is asked whether to personalize the sorting.
[0040] In another embodiment of the present application, items for sale similar to the item to be searched are retrieved in a target area, and the user is asked whether a personalized display is desired, specifically including: obtaining the user's historical search records, purchase records, and collection records, as well as the user's feedback, and using a machine learning algorithm to construct a user portrait; asking the user whether personalized sorting is performed based on the user portrait, and asking whether to repeat the prompt.
[0041] In one embodiment of the present application, the method also includes: if the product category is not found, asking the user whether to specify a local area, extracting local features of the local area through an attention mechanism, and obtaining a first feature map; performing channel compression on the product features of the searched product to obtain two second feature maps containing spatial information; splicing the second feature maps, and performing feature extraction and fusion through a convolutional layer to obtain a 2-channel third feature map; converting the third feature map into a spatial attention weight map, and multiplying it element-by-element with the first feature map to relax or change the product features.
[0042] In one embodiment of the present application, the method further includes: when a user clicks on a local area, extracting local features of the local area through a small convolution module; and performing batch pre-classification on multiple product images to obtain multiple product categories.
[0043] The above is an embodiment of the method proposed in this application. Based on the same inventive concept, the embodiment of this application also provides a vector-based platform image search and display device, whose structure is as follows Figure 2 shown.
[0044] Figure 2 A schematic diagram of the internal structure of a vector-based platform image search and display device provided in an embodiment of the present application. Figure 2 As shown, the device includes: at least one processor 201; and, a memory 202 communicatively connected to the at least one processor; The memory 202 stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor 201 to enable at least one processor 201 to: Receiving, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein there is at least one product to be searched, and the product to be searched includes at least one product image or product link; Using a deep learning model, extract product features of the product to be searched, and represent the product features as a first feature vector; Pre-classify the first feature vector to obtain commodity categories of the commodity to be searched; Obtain a to-be-searched area according to the product type, and perform partition retrieval on the to-be-searched products through the to-be-searched area to obtain a target area; Retrieve products for sale in the target area that are similar to the product being searched, and ask the user whether to display them in a personalized way.
[0045] Some embodiments of the present application provide corresponding Figure 1 A non-volatile computer storage medium for vector-based platform image search and display, storing computer executable instructions, wherein the computer executable instructions are set as: Receiving, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein there is at least one product to be searched, and the product to be searched includes at least one product image or product link; Using a deep learning model, extract product features of the product to be searched, and represent the product features as a first feature vector; Pre-classify the first feature vector to obtain commodity categories of the commodity to be searched; Obtain a to-be-searched area according to the product type, and perform partition retrieval on the to-be-searched products through the to-be-searched area to obtain a target area; Retrieve products for sale in the target area that are similar to the product being searched, and ask the user whether to display them in a personalized way.
[0046] Each embodiment in this application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the IoT device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0047] The system and medium provided in the embodiments of the present application correspond one-to-one to the method. Therefore, the system and medium also have similar beneficial technical effects to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the system and medium will not be repeated here.
[0048] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0049] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0050] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0051] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0052] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0053] Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
[0054] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0055] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0056] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. A vector-based platform image search and display method, characterized in that: The method comprises: Receiving, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein the product to be searched includes at least one product image or a product link; Using a deep learning model, extracting product features of the product to be searched, and representing the product features as a first feature vector; Pre-classifying the first feature vector to obtain commodity categories of the commodity to be searched; Obtaining a to-be-searched area according to the commodity type, and performing a partition search on the to-be-searched commodity through the to-be-searched area to obtain a target area; Retrieve the products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized way.
2. A vector-based platform image search and display method according to claim 1, characterized in that: Before receiving the to-be-searched product uploaded by the user through the e-commerce platform interface, the method further includes: The image and description text of the product for sale are converted into a high-dimensional vector to obtain a second feature vector; The second feature vector forms data points in a vector space, each data point corresponding to an image and description text of a product for sale; According to the similarity between the data points, a star map structure unit is constructed, wherein the connection between the data points represents the similarity relationship between the images; In the star map structure unit, a first preset number of data points are selected as starting points, and a second preset number of data points are expanded in all directions according to the similarity relationship, and are aggregated together to obtain the commodity category.
3. A vector-based platform image search and display method according to claim 2, characterized in that: The method includes obtaining a search area according to the product category, and performing a partition search on the product to be searched through the search area to obtain a target area. Specifically, the method includes: Selecting data points in the middle of the area to be searched as entry points respectively, and gradually expanding to surrounding data points along the similarity relationship; Using storage and calculation asynchronous retrieval, reading the image and calculating the distance at the same time, calculating the similarity between the first feature vector and the second feature vector; The data point with the largest similarity is selected as the target area.
4. A vector-based platform image search and display method according to claim 1, characterized in that: Retrieving products for sale similar to the product to be searched in the target area, and asking the user whether to display them in a personalized manner, specifically including: Perform similarity calculation in the target area to obtain a preset number of first search results; When the user selects personalized sorting, the search results are dynamically adjusted to display the second search result; If the user chooses to repeat the prompt, ask the user whether to personalize the sorting.
5. A vector-based platform image search and display method according to claim 4, characterized in that: Retrieving products for sale similar to the product to be searched in the target area, and asking the user whether to display them in a personalized manner, specifically including: Obtain the user's historical search history, purchase history, and collection history, as well as user feedback, and use machine learning algorithms to build user portraits; The user is asked whether to perform personalized sorting according to the user portrait, and whether to repeat the prompt.
6. A vector-based platform image search and display method according to claim 1, characterized in that: The method further comprises: If the product category is not found, the user is asked whether to specify a local area, and local features of the local area are extracted through an attention mechanism to obtain a first feature map; Channel compression is performed on the product features of the searched product to obtain two second feature maps containing spatial information; The second feature maps are concatenated, and feature extraction and fusion are performed through a convolutional layer to obtain a third feature map of 2 channels; The third feature map is converted into a spatial attention weight map and is element-wise multiplied with the first feature map to relax or change the product features.
7. A vector-based platform image search and display method according to claim 6, characterized in that: The method further comprises: When the user clicks on a local area, a local feature of the local area is extracted through a small convolution module; For multiple product images, batch pre-classification is performed to obtain multiple product categories.
8. The vector-based e-commerce platform image search method according to claim 1, characterized in that: Receive the products to be searched uploaded by users through the e-commerce platform interface, including: Scan the product to be searched in the ES server and the picture library through the picture scanning subsystem; Write the image data into the preset directory through the image writing subsystem; The image download subsystem downloads images on a large scale from the preset directory, performs preprocessing, and generates requests for image feature extraction tasks.
9. A vector-based platform image search and display device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to: Receiving, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein the product to be searched includes at least one product image or a product link; Using a deep learning model, extracting product features of the product to be searched, and representing the product features as a first feature vector; Pre-classifying the first feature vectors to obtain commodity categories of the commodities to be searched; Obtaining a to-be-searched area according to the commodity type, and performing a partition search on the to-be-searched commodity through the to-be-searched area to obtain a target area; Retrieve the products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized way.
10. A non-volatile computer storage medium for vector-based platform image search and display, storing computer executable instructions, characterized in that: The computer executable instructions are configured to: Receiving, through an e-commerce platform interface, a product to be searched uploaded by a user, wherein the product to be searched includes at least one product image or a product link; Using a deep learning model, extracting product features of the product to be searched, and representing the product features as a first feature vector; Pre-classifying the first feature vectors to obtain commodity categories of the commodities to be searched; Obtaining a to-be-searched area according to the commodity type, and performing a partition search on the to-be-searched commodity through the to-be-searched area to obtain a target area; Retrieve the products for sale similar to the product to be searched in the target area, and ask the user whether to display them in a personalized way.