Product resource evaluation method and device, equipment, storage medium and program product

Abstract

The invention relates to a product resource evaluation method and device, computer equipment, a storage medium and a program product. The method comprises the steps of extracting product image features of a to-be-detected target product, performing flaw detection on the product image features to obtain a flaw detection result, and performing product color recognition on the target product according to product attribute features corresponding to the target product and the flaw detection result to obtain a color recognition result. And collecting product market data associated with the target product, performing initial resource evaluation on the target product according to the defect detection result, the color recognition result and the product market data, obtaining an initial resource evaluation result, determining a pricing adjustment strategy corresponding to the target product, and adjusting the initial resource evaluation result according to the pricing adjustment strategy, and obtaining a resource evaluation result corresponding to the target product. By adopting the method, the accuracy of a resource evaluation result can be improved, and the resource interaction success rate of each product on the platform and the operation efficiency of the platform are improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular, to a product resource evaluation method, device, computer device, computer-readable storage medium, and computer program product. Background Art

[0002] With the development of computer technology and the popularization and application of various resource interaction platforms or application programs, etc., before releasing products on different platforms or application programs, it is necessary to conduct resource evaluation and pricing for different categories of products respectively, so as to improve the resource interaction success rate of different products and the operation efficiency of the platform through accurate pricing.

[0003] Traditionally, usually by collecting the historical evaluation results and historical pricing of the products to be released on different platforms or application programs, and based on the historical evaluation results and historical pricing for data analysis and prediction to obtain the current evaluation result and current pricing corresponding to the products to be released, it is possible to comprehensively consider the historical data on different platforms and improve the accuracy of the determined current evaluation result and current pricing.

[0004] However, the traditional way of obtaining the current evaluation result and current pricing of products, because it usually only considers the historical evaluation results of the products and does not further combine the performance and attributes of the products themselves, there are still problems of incomplete evaluation data, resulting in low accuracy of resource evaluation and pricing results. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide a product resource evaluation method, device, computer device, computer-readable storage medium, and computer program product that can improve the accuracy of resource evaluation and pricing results for products.

[0006] In a first aspect, this application provides a product resource evaluation method, including: extracting the product image features of a target product to be detected, performing defect detection on the product image features to obtain a defect detection result corresponding to the target product; obtaining the product attribute features corresponding to the target product, and based on the product attribute features and the defect detection result, performing product color identification on the target product to obtain a color identification result corresponding to the target product; collecting product market data associated with the target product, and based on the defect detection result, the color identification result, and the product market data, performing an initial resource evaluation on the target product to obtain an initial resource evaluation result corresponding to the target product; determining a pricing adjustment strategy corresponding to the target product, and based on the pricing adjustment strategy, adjusting the initial resource evaluation result of the target product to obtain a resource evaluation result corresponding to the target product.

[0007] In a second aspect, the present application further provides a product resource evaluation device, including: a defect detection module, configured to extract product image features of a target product to be detected, perform defect detection on the product image features, and obtain a defect detection result corresponding to the target product; a product quality identification module, configured to acquire product attribute features corresponding to the target product, and perform product quality identification on the target product according to the product attribute features and the defect detection result, so as to obtain a quality identification result corresponding to the target product; an initial resource evaluation module, configured to collect product market data associated with the target product, and perform an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, so as to obtain an initial resource evaluation result corresponding to the target product; a resource evaluation result obtaining module, configured to determine a pricing adjustment strategy corresponding to the target product, and adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy, so as to obtain a resource evaluation result corresponding to the target product.

[0008] In a third aspect, the present application further provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the following steps are implemented: extracting product image features of a target product to be detected, performing defect detection on the product image features, and obtaining a defect detection result corresponding to the target product; acquiring product attribute features corresponding to the target product, performing product quality identification on the target product according to the product attribute features and the defect detection result, and obtaining a quality identification result corresponding to the target product; collecting product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, so as to obtain an initial resource evaluation result corresponding to the target product; determining a pricing adjustment strategy corresponding to the target product, and adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, so as to obtain a resource evaluation result corresponding to the target product.

[0009] Fourthly, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: extracting product image features of a target product to be detected, performing defect detection on the product image features to obtain a defect detection result corresponding to the target product; acquiring product attribute features corresponding to the target product, and performing product quality identification on the target product according to the product attribute features and the defect detection result to obtain a quality identification result corresponding to the target product; collecting product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data to obtain an initial resource evaluation result corresponding to the target product; determining a pricing adjustment strategy corresponding to the target product, and adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain a resource evaluation result corresponding to the target product.

[0010] Fifthly, the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented: extracting product image features of a target product to be detected, performing defect detection on the product image features to obtain a defect detection result corresponding to the target product; acquiring product attribute features corresponding to the target product, and performing product quality identification on the target product according to the product attribute features and the defect detection result to obtain a quality identification result corresponding to the target product; collecting product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data to obtain an initial resource evaluation result corresponding to the target product; determining a pricing adjustment strategy corresponding to the target product, and adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain a resource evaluation result corresponding to the target product.

[0011] In the above product resource evaluation method, device, computer device, computer-readable storage medium, and computer program product, by extracting the product image features of the target product to be detected, performing defect detection on the product image features to obtain a defect detection result corresponding to the target product, and obtaining the product attribute features corresponding to the target product, so as to identify the product quality of the target product according to the product attribute features and the defect detection result, and obtain a quality identification result corresponding to the target product. Thus, a comprehensive and all-round evaluation of the target product can be realized from the attributes of the target product itself, including defects and product attributes, etc., reducing the error in the resource evaluation process. Further, by collecting product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, an initial resource evaluation result corresponding to the target product is obtained. Thus, a comprehensive evaluation of the target product can be carried out from multiple angles, improving the accuracy of the obtained initial resource evaluation result. And after obtaining the initial resource evaluation result, by determining a pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, a dynamic adjustment of the resource evaluation result of the target product is realized, improving the accuracy of the obtained resource evaluation result corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required to be used in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can be obtained based on these drawings.

[0013] Figure 1 It is an application environment diagram of the product resource evaluation method in an embodiment;

[0014] Figure 2 It is a flowchart of the product resource evaluation method in an embodiment;

[0015] Figure 3 It is a flowchart of determining a pricing adjustment strategy corresponding to the target product in an embodiment;

[0016] Figure 4 It is a flowchart of obtaining a quality identification result corresponding to the target product in an embodiment;

[0017] Figure 5 It is a flowchart of the product resource evaluation method in another embodiment;

[0018] Figure 6 It is a schematic flow chart of a product resource evaluation method in yet another embodiment;

[0019] Figure 7 It is a structural block diagram of a product resource evaluation device in an embodiment;

[0020] Figure 8 It is an internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0021] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0022] The product resource evaluation method provided by the embodiments of the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal device 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or on other network servers. Among them, the terminal device 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or on other network servers. Among them, the terminal device 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices, portable wearable devices, and aircraft, etc. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, and projection devices, etc. The portable wearable devices can be smart watches, smart bracelets, and head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, a smart glasses, etc. The server 104 can be an independent physical server, or a server cluster composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Among them, the terminal device 102 and the server 104 can be directly or indirectly connected through wired or wireless communication methods, and this is not limited in the embodiments of the present application.

[0023] Among them, both the terminal device 102 and the server 104 can be separately used to execute the product resource evaluation method provided in the embodiments of the present application, and the terminal device 102 and the server 104 can also cooperate to execute the product resource evaluation method provided in the embodiments of the present application. For example, taking the cooperation between the terminal device 102 and the server 104 to execute the product resource evaluation method provided in the embodiments of the present application as an example, the server 104 extracts the product image features of the target product to be detected, performs defect detection on the product image features, obtains a defect detection result corresponding to the target product, and obtains product attribute features corresponding to the target product, so as to identify the product quality of the target product according to the product attribute features and the defect detection result, and obtain a quality identification result corresponding to the target product. Further, the server 104 collects product market data associated with the target product, and performs an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, and obtains an initial resource evaluation result corresponding to the target product. Among them, the server 104 determines a pricing adjustment strategy corresponding to the target product, and adjusts the initial resource evaluation result of the target product according to the pricing adjustment strategy, obtains a resource evaluation result corresponding to the target product, and feeds back the obtained resource evaluation result to the terminal device 102 for display, for the user corresponding to the terminal device 102 to view and select.

[0024] In an exemplary embodiment, as Figure 2 shown, a product resource evaluation method is provided. This method can be executed separately by the terminal device or the server, or can be executed jointly by the terminal device and the server. Taking this method applied to Figure 1 the server 104 in as an example for description, it includes the following steps S202 to step S208. Among them:

[0025] Step S202, extract the product image features of the target product to be detected, perform defect detection on the product image features, and obtain a defect detection result corresponding to the target product.

[0026] Among them, a resource interaction platform or application program is running on the terminal device. The server can publish various different types of products based on the resource interaction platform or application program. Then, the user can trigger product viewing requests, access requests, purchase requests, etc. based on different products published by the resource interaction platform or application program. Before the server publishes different categories of products based on the resource interaction platform or application program, it needs to perform a resource evaluation on the products to be published to obtain a resource evaluation result, so as to determine the product pricing corresponding to the products to be published with the resource evaluation result.

[0027] Specifically, before the server publishes different categories of products based on the resource interaction platform or application, it obtains the product image of the target product to be published and extracts the product image features of the product image. Specifically, it can use a pre-trained vision large model, such as the multi-modal pre-trained model CLIP (Contrastive Language-Image Pre-training, which means a pre-trained model that embeds images and text into a shared semantic space through contrastive learning to achieve cross-modal understanding), and the prompt-based segmentation model SAM (Segment Anything Model, which means a model that can segment any object in various images through pre-training, even if these objects have never appeared in the training data), etc., to extract features from the product image and obtain the product image features corresponding to the target product.

[0028] Furthermore, the server uses the trained defect detection model to detect defects in the product image features and obtains the defect detection result corresponding to the target product. Specifically, the server uses the trained defect detection model to detect defects and locate defects in the product image features, obtains at least one defect corresponding to the product image features and the specific location of at least one defect on the target product, obtains the preset defect item configuration table, parses the preset defect item configuration table, obtains the mapping relationship between different defects and defect categories in the preset defect item configuration table, classifies at least one defect, obtains the defect category corresponding to each of at least one defect, and determines at least one defect and the defect category corresponding to each of at least one defect as the defect detection result corresponding to the target product.

[0029] Among them, the defects corresponding to the product image features can be used to represent the damage situation of the target product. According to the different defect severity levels corresponding to the defect categories, the degree of damage to the target product is different. Among them, the defect categories specifically include scratches, knocks, and cracks, etc., and the defect severity levels of scratches, knocks, and cracks increase in turn, that is, the defect severity level of scratches < the defect severity level of knocks < the defect severity level of cracks, that is, the degree of damage to the target product when there are scratches < the degree of damage to the target product when there are knocks < the degree of damage to the target product when there are cracks.

[0030] Exemplarily, the trained defect detection model specifically can be obtained by training and fine-tuning initial detection models such as the YOLO model (fully known as You Only Look Once, which is a deep learning model for real-time object detection. Its core idea is to transform the object detection problem into a regression problem and simultaneously predict multiple bounding boxes and class probabilities through a single forward pass) and the Mask R-CNN model (fully known as Mask Fast Region-based Convolutional Network, which can be understood as a deep learning model for object detection and instance segmentation. Its architecture is Fast Region-based Convolutional Network, with an additional branch for generating segmentation masks). Specifically, it can be obtained by collecting image sample data and annotating the image sample data, including annotating as defective and non-defective, etc., so as to train the initial detection model based on the annotated image sample data, and when the model training end condition is met, obtain the trained defect detection model.

[0031] Step S204: Obtain the product attribute features corresponding to the target product, and based on the product attribute features and the defect detection result, perform product quality identification on the target product to obtain the quality identification result corresponding to the target product.

[0032] Specifically, the server extracts features from the defect detection result to obtain the product defect features corresponding to the target product. At the same time, the server obtains the product attribute information corresponding to the target product, including information such as the brand, model, usage years, initial price, and release platform of the product, and extracts features from each product attribute information to obtain the product attribute features corresponding to the target product.

[0033] Among them, the server also needs to obtain the product evaluation information corresponding to the target product, perform semantic analysis and feature extraction on the product evaluation information to obtain the product evaluation features corresponding to the target product, so that the to-be-identified defect features corresponding to the target product can be obtained by combining the product attribute features, product evaluation features, and product defect features. Among them, the product evaluation information can specifically be the evaluations of different users on the target product, such as usage evaluations of the product, appearance evaluations of the product (whether there are defects, etc.), and pricing evaluations of the product, etc.

[0034] Furthermore, the server performs product quality identification on the to-be-identified defect features to obtain the quality identification result corresponding to the target product. Among them, the quality identification result is used to represent the newness or depreciation degree of the target product. For example, the quality identification result of the target product is 90% new, or 80% new, or 50% new, etc., with different newness degrees.

[0035] Step S206: Collect product market data associated with the target product, and based on the defect detection results, the user's evaluation information, the color identification results, and the product market data, conduct an initial resource assessment of the target product to obtain an initial resource assessment result corresponding to the target product.

[0036] Specifically, the server collects product market data associated with the target product, which specifically includes the historical transaction prices of the target product on different resource interaction platforms or applications, the evaluation information of different users on the target product, the product description information of the target product, etc., and performs data fusion on the defect detection results, color identification results, and product market data to obtain multi-modal product features corresponding to the target product.

[0037] Furthermore, the server uses the trained resource assessment model to perform product identification and resource assessment on the multi-modal product features, thereby obtaining an initial resource assessment result corresponding to the target product. Among them, by inputting the multi-modal product features into the trained resource assessment model, the resource assessment model can perform product identification and resource assessment on the multi-modal product features, including processing such as vectorizing text information and vectorizing image information of the multi-modal product features to obtain multi-modal feature vectors, and obtaining the initial resource assessment result through regression prediction of the multi-modal feature vectors. And due to training, the resource assessment model learns the complex non-linear relationships and the associations between multi-modal data, thereby improving the accuracy of the initial resource assessment result obtained by using the resource assessment model for resource assessment.

[0038] Exemplarily, the trained resource assessment model can specifically be obtained by training initial resource assessment models such as the deepseek-R1 model (referring to a super-large pre-trained language model, which is based on deep learning technology, adopts a mixture of experts architecture, and performs well in natural language processing tasks and can handle complex reasoning tasks), and the qwen-plus model (referring to a large pre-trained language model, which aims to provide efficient and accurate natural language understanding and generation capabilities, and this model is based on the Transformer architecture in deep learning technology and can handle various complex language tasks).

[0039] Among them, by collecting picture sample data of different products and annotating the picture sample data, including annotating the estimated price for the picture sample data with defects and annotating the estimated price for the picture sample data without defects, etc., a training sample set is obtained, and the training sample set is used to train the initial resource assessment model. During the training process, the model parameters of the initial resource assessment model are optimized and adjusted until the model training end condition is met, and a trained resource assessment model is obtained.

[0040] Step S208: Determine the pricing adjustment strategy corresponding to the target product, and adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain the resource evaluation result corresponding to the target product.

[0041] Specifically, the server determines the resource interaction platform or application program to which the target product belongs, and determines the pricing adjustment strategy set for the target product under this resource interaction platform or application program, so as to dynamically adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy, and obtain the resource evaluation result for the target product under different circumstances.

[0042] In the above product resource evaluation method, by extracting the product image features of the target product to be detected, detecting the defects of the product image features to obtain the defect detection result corresponding to the target product, and obtaining the product attribute features corresponding to the target product, so as to identify the product quality of the target product according to the product attribute features and the defect detection result, and obtain the quality identification result corresponding to the target product. Thus, a comprehensive and all-round evaluation of the target product can be realized from the attributes of the target product itself, including defects and product attributes, etc., reducing the error in the resource evaluation process. Further, by collecting the product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result and the product market data, the initial resource evaluation result corresponding to the target product is obtained. Thus, a comprehensive evaluation of the target product can be carried out from multiple angles, improving the accuracy of the obtained initial resource evaluation result. And after obtaining the initial resource evaluation result, by determining the pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, the dynamic adjustment of the resource evaluation result of the target product is realized, improving the accuracy of the obtained resource evaluation result corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform.

[0043] In an exemplary embodiment, as Figure 3 shown, the steps of determining the pricing adjustment strategy corresponding to the target product include the following steps S302 to S310. Among them:

[0044] Step S302: Determine the resource interaction platform to which the target product belongs.

[0045] Specifically, the server determines the resource interaction platform to which the target product belongs by obtaining the product attribute information corresponding to the target product, including information such as the brand, model, service life, initial price, and release platform of the product.

[0046] Step S304: Obtain object behavior data associated with the target product based on the resource interaction platform.

[0047] Specifically, after determining the resource interaction platform for releasing the target product, further obtain the object behavior data triggered by users for the target product on the resource interaction platform. For example, whether users click on the target product, the time when different users click on the target product, and the number of times different users click on the target product. It also includes data such as the quotes made by users for the target product and the transaction price of the target product when a deal is reached.

[0048] Step S306: Determine the resource interaction status data corresponding to the target product based on the product attribute characteristics, object behavior data, and product market data.

[0049] Specifically, the server constructs a Markov Decision Process (i.e., MDP, full name Markov Decision Process, understood as a mathematical framework for modeling decision problems of dynamic systems with randomness, applied in fields such as reinforcement learning, operations research, economics, and engineering, used to optimize decision sequences to achieve the desired goal) model based on the resource interaction platform, and determines the resource interaction status data for the Markov Decision Process model based on the product attribute characteristics, object behavior data, and product market data, as the resource interaction status data corresponding to the target product.

[0050] Exemplarily, the goal of the Markov Decision Process is to find an optimal policy π∗ such that starting from any initial state, when acting according to this policy, the expected value of the cumulative reward is maximized. Its core ideas include: 1) State: The state represents the situation of the system at a certain moment, and the state space S is the set of all possible states. 2) Action: The action is the behavior that the decision maker can take in a certain state, and the action space A is the set of all possible actions. 3) Transition Probability: The transition probability P(s′∣s,a) represents the probability that the system transfers to state s′ after taking action a in state s. 4) Reward: The reward R(s,a) is the immediate return obtained after taking action a in state s, and the reward function defines the immediate value of each state-action pair. 5) Policy: The policy π(a∣s) is the probability distribution of choosing action a in state s, and the policy defines the behavior mode of the decision maker. 6) Discount Factor: The discount factor γ (0 ≤ γ ≤ 1) is used to measure the current value of future rewards, which reflects the degree of importance the decision maker attaches to future rewards.

[0051] Furthermore, specifically in the embodiments of the present application, when constructing a Markov decision process model based on the resource interaction platform, the resource interaction state data includes: 1) State, specifically including product attribute features (referring to the information description of the commodity, including information such as the brand, model, service life, initial pricing, and release platform of the product), product market data (specifically including the market supply and demand relationship for the target product), and object behavior data (whether the user clicks on the target product, the time when different users click on the target product, and the number of times different users click on the target product, and also including data such as the user's quote for the target product and the transaction price of the target product when a deal is reached). 2) Action, specifically represented as a price adjustment strategy in space (based on the premise that a deal can be reached after adjusting the product's pricing, the adjustment strategy is mainly based on two situations: sufficient exposure - no quote and sufficient quote - no deal. The reference price is lowered, and the intensity is adjusted based on the results given by reinforcement learning until a deal is made). 3) Reward function, specifically designed based on indicators such as transaction success rate, transaction time, and user satisfaction. Thus, through interaction with the resource interaction platform, the optimal price adjustment strategy of reinforcement learning is obtained to improve the transaction efficiency of the platform.

[0052] Step S308: Based on the resource interaction platform, obtain the resource interaction records associated with the target product, and determine the reward signal parameters corresponding to the target product according to the resource interaction records and the object behavior data.

[0053] Specifically, the server further obtains the resource interaction records associated with the target product based on the resource interaction platform, specifically including data such as the historical quotes of different users for the target product and the historical transaction price of the target product when a deal is reached. Thus, according to the resource interaction records (including the historical quotes of different users for the target product and the historical transaction price of the target product when a deal is reached), and the object behavior data (including whether the user clicks on the target product, the time when different users click on the target product, and the number of times different users click on the target product, and also including data such as the user's quote for the target product and the transaction price of the target product when a deal is reached), combined with the setting rules of the reward signal parameters (including setting according to indicators such as transaction success rate, transaction time, and user satisfaction), the reward signal parameters corresponding to the target product are determined.

[0054] Step S310: Determine the pricing adjustment strategy corresponding to the target product according to the resource interaction state data and the reward signal parameters.

[0055] Specifically, in the Markov decision process, the action is specifically represented as a price adjustment strategy in space. Based on the premise that a deal can be reached after adjusting the pricing of the product, the adjustment strategy is mainly carried out based on two situations: sufficient exposure - no quotation and sufficient quotation - no deal. The reference price is lowered, and the intensity is adjusted according to the result given by reinforcement learning until a deal is made. Therefore, when adjusting the initial resource evaluation result of the target product, the pricing adjustment strategy for the target product can be determined according to the resource interaction status data and reward signal parameters corresponding to the target product, and the specific adjustment intensity can be determined according to the pricing adjustment strategy. Then, the initial resource evaluation result is adjusted according to the specific adjustment intensity to obtain the resource evaluation result corresponding to the target product. Specifically, the initial resource evaluation result (i.e., the initial pricing of the target product) is adjusted to obtain the adjusted resource evaluation result (i.e., the actual product pricing of the target product).

[0056] In this embodiment, by determining the resource interaction platform to which the target product belongs, based on the resource interaction platform, the object behavior data associated with the target product and the resource interaction record are obtained. According to the product attribute characteristics, object behavior data, and product market data, the resource interaction status data corresponding to the target product is determined. Thus, the reward signal parameters corresponding to the target product can be determined according to the resource interaction record and object behavior data, and the pricing adjustment strategy corresponding to the target product can be determined according to the resource interaction status data and reward signal parameters. Therefore, the initial resource evaluation result of the target product can be adjusted according to the pricing adjustment strategy, realizing the dynamic adjustment of the resource evaluation result of the target product, improving the accuracy of the obtained resource evaluation result corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform.

[0057] In an exemplary embodiment, as Figure 4 shown, the steps of obtaining the color recognition result corresponding to the target product, that is, the steps of performing product color recognition on the target product according to the product attribute characteristics and the defect detection result to obtain the color recognition result corresponding to the target product, specifically include the following steps S402 to S408. Among them:

[0058] Step S402, extract the features of the defect detection result to obtain the product defect features corresponding to the target product.

[0059] Specifically, the server performs semantic recognition and feature extraction on the defect detection results to obtain product defect features corresponding to the target product, which may specifically include the number of defects and the categories of defects. The defect categories specifically include scratches, knocks, and cracks, etc. The severity levels of scratches, knocks, and cracks increase in sequence, that is, the severity level of scratches < the severity level of knocks < the severity level of cracks, which means the damage degree of the target product when scratches appear < the damage degree of the target product when knocks appear < the damage degree of the target product when cracks appear.

[0060] Step S404: Obtain product evaluation information corresponding to the target product, perform semantic analysis and feature extraction on the product evaluation information, and obtain product evaluation features corresponding to the target product.

[0061] Specifically, the server obtains product evaluation information associated with the target product based on the resource interaction platform that releases the target product. Specifically, it can be the evaluations of different users on the target product, such as usage evaluations of the product, appearance evaluations of the product (whether there are defects, etc.), and pricing evaluations of the product, etc. And perform semantic analysis and feature extraction on the product evaluation information to obtain product evaluation features corresponding to the target product.

[0062] Step S406: Combine the product attribute features, product evaluation features, and product defect features to obtain the defect features to be identified corresponding to the target product.

[0063] Specifically, the server combines the product attribute features, product evaluation features, and product defect features, that is, from multiple aspects such as the attributes of the product itself, the evaluations of users on the product, and the defects existing in the product, performs data fusion on the product attribute features, product evaluation features, and product defect features, and obtains the defect features to be identified corresponding to the target product.

[0064] Step S408: Obtain the color and configuration information associated with the target product, and according to the color and configuration information, perform product color identification on the defect features to be identified, and obtain the color identification result corresponding to the target product.

[0065] Specifically, the server obtains the color and configuration information associated with the target product, and parses the color and configuration information to obtain the mapping relationship between the defect features and the color identification result in the color and configuration information, so as to perform product color identification on the defect features to be identified according to the mapping relationship between the defect features and the color identification result, and determine the color identification result corresponding to the defect features to be identified.

[0066] Among them, the color identification result is used to represent the newness or damage degree of the target product. For example, the color identification result of the target product is 90% new, or 80% new, or 50% new, etc. with different newness degrees. That is, in the color configuration information, for different defect features, corresponding color identification results are respectively configured.

[0067] Exemplarily, the defect categories in the defect features specifically include scratches, knocks, and cracks, etc., and the severity levels of scratches, knocks, and cracks increase in sequence. Then, for different defect categories in the defect features, different color identification results such as 90% new, or 80% new, or 50% new, etc. are respectively set. For example, specifically, the mapping relationship can be that scratches correspond to 90% new, knocks correspond to 80% new, and cracks correspond to 50% new, etc.

[0068] In this embodiment, by extracting features from the defect detection result, the product defect features corresponding to the target product are obtained, and the product evaluation information corresponding to the target product is acquired. Through semantic analysis and feature extraction of the product evaluation information, the product evaluation features corresponding to the target product are obtained, and the product attribute features, product evaluation features, and product defect features are combined. Thus, by comprehensively considering multiple different aspects, the defect features to be identified corresponding to the target product are obtained comprehensively, reducing feature omission and error data in the identification process. Further, by obtaining the color configuration information associated with the target product, the product color can be identified for the defect features to be identified according to the color configuration information, and the color identification result corresponding to the target product can be obtained quickly and accurately, improving the working efficiency of the color identification processing process.

[0069] In an exemplary embodiment, as Figure 5 shown, a product resource evaluation method is provided. Taking the method applied to Figure 1 the server 104 therein as an example for illustration, it includes the following steps S502 to step S514. Among them:

[0070] Step S502, extract the product image features of the target product to be detected, and perform defect detection on the product image features to obtain the defect detection result corresponding to the target product.

[0071] Specifically, before the server publishes different categories of products based on the resource interaction platform or application program, by obtaining the product image of the target product to be published and extracting the product image features of the product image, specifically, the product image features corresponding to the target product can be obtained by using a pre-trained visual large model to extract features from the product image.

[0072] Further, the server uses the trained defect detection model to detect defects in the product image features and obtains the defect detection result corresponding to the target product. Specifically, the server uses the trained defect detection model to detect and locate defects in the product image features, obtains at least one defect corresponding to the product image features and the specific location of at least one defect on the target product, and by obtaining a preset defect item configuration table and parsing the preset defect item configuration table, obtains the mapping relationship between different defects and defect categories in the preset defect item configuration table, classifies at least one defect, obtains the defect category corresponding to each of at least one defect, and determines at least one defect and the defect category corresponding to each of at least one defect as the defect detection result corresponding to the target product.

[0073] Step S504: Obtain the product attribute features corresponding to the target product, and identify the product condition of the target product according to the product attribute features and the defect detection result, and obtain the condition identification result corresponding to the target product.

[0074] Specifically, the server extracts the product defect features corresponding to the target product by extracting features from the defect detection result. At the same time, the server obtains the product attribute information corresponding to the target product, including information such as the brand, model, service life, initial price, and release platform of the product, and extracts features from each product attribute information to obtain the product attribute features corresponding to the target product. Among them, the server also needs to obtain the product evaluation information corresponding to the target product, and perform semantic analysis and feature extraction on the product evaluation information to obtain the product evaluation features corresponding to the target product, so as to obtain the defect features to be identified corresponding to the target product by combining the product attribute features, product evaluation features, and product defect features. The product evaluation information may specifically be the evaluations of different users on the target product, such as usage evaluations of the product, appearance evaluations of the product (whether there are defects, etc.), and pricing evaluations of the product.

[0075] Further, the server obtains the condition configuration information associated with the target product and parses the condition configuration information to obtain the mapping relationship between the defect features and the condition identification result in the condition configuration information, so as to identify the product condition of the defect features to be identified according to the mapping relationship between the defect features and the condition identification result, and determine the condition identification result corresponding to the defect features to be identified. The condition identification result is used to represent the degree of newness or depreciation of the target product. For example, the condition identification result of the target product is 90% new, or 80% new, or 50% new, etc.

[0076] Step S506: Collect product market data associated with the target product, perform data fusion on the defect detection results, color identification results, and product market data to obtain multi-modal product features corresponding to the target product.

[0077] Specifically, the server collects product market data associated with the target product, including the historical transaction prices of the target product on different resource interaction platforms or applications, the evaluation information of different users on the target product, the product description information of the target product, etc., and performs data fusion on the defect detection results, color identification results, and product market data to obtain multi-modal product features corresponding to the target product.

[0078] Step S508: Collect labeled sample data carrying defect annotations and historical resource evaluation results, and unsupervised learning data carrying product evaluation information.

[0079] Specifically, the server collects picture sample data of different products and annotates the picture sample data, including annotating the estimated prices for picture sample data with defects and annotating the estimated prices for picture sample data without defects, etc., to obtain a training sample set.

[0080] Step S510: Train the initial resource evaluation model according to the labeled sample data and the unsupervised learning data, and fine-tune the initial resource evaluation model during the model training process according to the reward signal parameters corresponding to the target product. When the model training end condition is met, obtain the trained resource evaluation model.

[0081] Specifically, the server uses the training sample set, including the labeled sample data and the unsupervised learning data, etc., to train the initial resource evaluation model. During the training process, the model parameters of the initial resource evaluation model are optimized and adjusted until the model training end condition is met, and the trained resource evaluation model is obtained.

[0082] Step S512: Perform product identification and resource evaluation on the multi-modal product features according to the trained resource evaluation model to obtain the initial resource evaluation result corresponding to the target product.

[0083] Specifically, the server uses the trained resource evaluation model to perform product identification and resource evaluation on the multi-modal product features, so as to obtain the initial resource evaluation result corresponding to the target product.

[0084] Among them, by inputting multi-modal product features into a trained resource evaluation model, the resource evaluation model can perform product identification and resource evaluation on the multi-modal product features, including processing such as vectorizing text information and vectorizing picture information of the multi-modal product features to obtain multi-modal feature vectors, and obtaining an initial resource evaluation result through regression prediction of the multi-modal feature vectors.

[0085] Step S514, determine a pricing adjustment strategy corresponding to the target product, and adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain a resource evaluation result corresponding to the target product.

[0086] Specifically, the server determines the resource interaction platform or application program to which the target product belongs, and determines the pricing adjustment strategy set for the target product under this resource interaction platform or application program, so as to dynamically adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy and obtain the resource evaluation result for the target product in different situations.

[0087] Among them, the server obtains the product attribute information corresponding to the target product, determines the resource interaction platform to which the target product belongs according to the release platform in the product attribute information, further obtains the object behavior data triggered by the user for the target product on the resource interaction platform, and the resource interaction record associated with the target product, so as to determine the resource interaction status data corresponding to the target product according to the product attribute characteristics, object behavior data, and product market data.

[0088] Furthermore, the server determines the reward signal parameter corresponding to the target product according to the resource interaction record (including the historical quotes of different users for the target product and the historical transaction pricing of the target product when a transaction is reached), and the object behavior data (including whether the user clicks on the target product, the time when different users click on the target product, and the number of times different users click on the target product, and also including data such as the quotes of the user for the target product and the transaction pricing of the target product when a transaction is reached), combined with the setting rules of the reward signal parameter (including setting according to indicators such as transaction success rate, transaction time, and user satisfaction). Thus, when adjusting the initial resource evaluation result of the target product, the pricing adjustment strategy for the target product can be determined according to the resource interaction status data and the reward signal parameter corresponding to the target product, and the specific adjustment strength can be determined according to the pricing adjustment strategy, so as to adjust the initial resource evaluation result according to the specific adjustment strength to obtain the resource evaluation result corresponding to the target product. Among them, specifically, the initial resource evaluation result (i.e., the initial pricing of the target product) is adjusted to obtain the adjusted resource evaluation result (i.e., the actual product pricing of the target product).

[0089] In the above product resource evaluation method, by extracting the product image features of the target product to be detected, detecting defects in the product image features, obtaining a defect detection result corresponding to the target product, and acquiring the product attribute features corresponding to the target product, so as to identify the product quality of the target product according to the product attribute features and the defect detection result, and obtain a quality identification result corresponding to the target product. Thus, a comprehensive and all-round evaluation of the target product can be realized from the attributes of the target product itself, including defects and product attributes, etc., reducing the error in the resource evaluation process. Further, by collecting product market data associated with the target product, and based on the defect detection result, the quality identification result, and the product market data, an initial resource evaluation of the target product is performed to obtain an initial resource evaluation result corresponding to the target product. Thus, a comprehensive evaluation of the target product can be carried out from multiple angles, improving the accuracy of the obtained initial resource evaluation result. And after obtaining the initial resource evaluation result, by determining a pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, a dynamic adjustment of the resource evaluation result of the target product is realized, improving the accuracy of the obtained resource evaluation result corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform.

[0090] In an exemplary embodiment, as Figure 6 shown, a product resource evaluation method is provided. Taking the case where this method is applied to the Figure 1 server 104 as an example, it includes the following steps S601 to step S615. Among them:

[0091] Step S601: Extract the product image features of the target product to be detected, detect defects in the product image features, and obtain at least one defect corresponding to the product image features.

[0092] Step S602: According to a preset defect item configuration table, classify the at least one defect to obtain a defect category corresponding to each of the at least one defect, and determine the at least one defect and the defect category corresponding to each of the at least one defect as the defect detection result corresponding to the target product.

[0093] Step S603: Extract features from the defect detection result to obtain product defect features corresponding to the target product.

[0094] Step S604: Obtain product evaluation information corresponding to the target product, perform semantic analysis and feature extraction on the product evaluation information, and obtain product evaluation features corresponding to the target product.

[0095] Step S605, obtaining product attribute features corresponding to the target product, combining product attribute features, product evaluation features and product defect features, and obtaining to-be-identified defect features corresponding to the target product.

[0096] Step S606, obtaining color configuration information associated with the target product, and performing product color recognition on the defect features to be recognized based on the color configuration information to obtain a color recognition result corresponding to the target product.

[0097] Step S607, collecting product market data associated with the target product, performing data fusion on the defect detection results, color recognition results and product market data, and obtaining multimodal product features corresponding to the target product.

[0098] Step S608, collecting labeled sample data carrying defect annotations and historical resource evaluation results, and unsupervised learning data carrying product evaluation information.

[0099] Step S609, training the initial resource evaluation model based on the labeled sample data and the unsupervised learning data, and fine-tuning the initial resource evaluation model during the model training process based on the reward signal parameters corresponding to the target product, and obtaining a trained resource evaluation model when the model training end conditions are met.

[0100] Step S610, performing product identification and resource evaluation on the multimodal product features according to the trained resource evaluation model, and obtaining an initial resource evaluation result corresponding to the target product.

[0101] Step S611, determining the resource interaction platform to which the target product belongs, and acquiring object behavior data associated with the target product based on the resource interaction platform.

[0102] Step S612, determining resource interaction state data corresponding to the target product according to product attribute characteristics, object behavior data and product market data.

[0103] Step S613, based on the resource interaction platform, obtaining the resource interaction record associated with the target product, and determining the reward signal parameter corresponding to the target product according to the resource interaction record and the object behavior data.

[0104] Step S614: determining a pricing adjustment strategy corresponding to the target product according to the resource interaction state data and the reward signal parameters.

[0105] Step S615: adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain a resource evaluation result corresponding to the target product.

[0106] In an exemplary embodiment, a product resource evaluation method is provided, which specifically includes: 1. Data collection: Collect relevant market data from multiple channels and perform data cleaning and preprocessing. 2. Defect detection: Use the defect detection module to analyze product images and automatically detect defect items. 3. Condition matching: According to the defect detection results and product characteristics, use the condition matching module to automatically match the product condition. 4. Initial price estimation: Combine the defect detection results, condition matching results, and market data, and use the price estimation module to perform initial price estimation. 5. Dynamic price adjustment: Use the reinforcement learning module to dynamically adjust the price according to the state and reward signal of the platform trading environment to optimize the platform trading efficiency. 6. Model training and optimization: Train and optimize the large model and reinforcement learning model to improve the accuracy and adaptability of the model. 7. User interaction: Through the user interaction module, users can upload product information and obtain pricing results, and at the same time provide feedback on the pricing results.

[0107] Among them, the specific effects that the above product resource evaluation method can achieve include: 1. Precise pricing: Through the powerful capabilities of the large model, it can make full use of the information in the massive data, learn complex non-linear relationships and the associations between multi-modal data, so as to achieve precise price estimation. 2. Dynamic adjustment: Combined with the reinforcement learning module, it can adjust the price in real time according to the dynamic changes of the platform trading environment to optimize the trading efficiency. 3. High degree of automation: The defect detection and condition matching modules can be completed automatically without manual intervention, improving the efficiency and accuracy of pricing. 4. Strong adaptability: Use reinforcement learning technology to optimize the model, enabling it to better adapt to market changes and user needs. 5. Good user experience: Users can conveniently upload product information and obtain pricing results through the user interaction module, and at the same time provide feedback on the pricing results, improving user satisfaction. 6. Wide application: It is applicable to product pricing in the second-hand market, as well as in multiple fields such as new product pricing, inventory management, and market prediction, with broad application prospects.

[0108] In the above product resource evaluation method, by extracting the product image features of the target product to be detected, performing defect detection on the product image features, obtaining the defect detection result corresponding to the target product, and acquiring the product attribute features corresponding to the target product, so as to identify the product quality of the target product according to the product attribute features and the defect detection result, and obtain the quality identification result corresponding to the target product. Thus, a comprehensive and overall evaluation of the target product can be realized from the attributes of the target product itself, including defects and product attributes, etc., reducing the error in the resource evaluation process. Further, by collecting the product market data associated with the target product, and based on the defect detection result, the quality identification result, and the product market data, performing an initial resource evaluation on the target product, and obtaining the initial resource evaluation result corresponding to the target product. Thus, a comprehensive evaluation of the target product can be carried out from multiple perspectives, improving the accuracy of the obtained initial resource evaluation result. And after obtaining the initial resource evaluation result, by determining the pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, the dynamic adjustment of the resource evaluation result of the target product is realized, improving the accuracy of the obtained resource evaluation result corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform.

[0109] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these steps or stages does not necessarily have to be sequential, but can be executed alternately or in turn with at least some of the steps or stages in other steps or other steps.

[0110] Based on the same inventive concept, an embodiment of the present application further provides a product resource evaluation device for implementing the above-mentioned product resource evaluation method. The solution provided by this device for solving problems is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the product resource evaluation device provided below can refer to the limitations on the product resource evaluation method in the above text, and will not be repeated here.

[0111] In an exemplary embodiment, as Figure 7As shown in the figure, a product resource evaluation device is provided, including: a defect detection module 702, a product quality identification module 704, an initial resource evaluation module 706, and a resource evaluation result obtaining module 708, where:

[0112] The defect detection module 702 is configured to extract the product image features of the target product to be detected, perform defect detection on the product image features, and obtain a defect detection result corresponding to the target product; the product quality identification module 704 is configured to obtain the product attribute features corresponding to the target product, and perform product quality identification on the target product according to the product attribute features and the defect detection result, so as to obtain a quality identification result corresponding to the target product; the initial resource evaluation module 706 is configured to collect the product market data associated with the target product, and perform an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, so as to obtain an initial resource evaluation result corresponding to the target product; the resource evaluation result obtaining module 708 is configured to determine a pricing adjustment strategy corresponding to the target product, and adjust the initial resource evaluation result of the target product according to the pricing adjustment strategy, so as to obtain a resource evaluation result corresponding to the target product.

[0113] In the above product resource evaluation device, by extracting the product image features of the target product to be detected, performing defect detection on the product image features, obtaining a defect detection result corresponding to the target product, and obtaining the product attribute features corresponding to the target product, so as to perform product quality identification on the target product according to the product attribute features and the defect detection result, and obtain a quality identification result corresponding to the target product, it is possible to realize a comprehensive and thorough evaluation of the target product from the attributes of the target product itself, including defects and product attributes, etc., and reduce the error in the resource evaluation process. Further, by collecting the product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the quality identification result, and the product market data, an initial resource evaluation result corresponding to the target product is obtained, so as to comprehensively evaluate the target product from multiple angles and improve the accuracy of the obtained initial resource evaluation result. And after obtaining the initial resource evaluation result, by determining a pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy, the dynamic adjustment of the resource evaluation result of the target product is realized, the accuracy of the obtained resource evaluation result corresponding to the target product is improved, and the resource interaction success rate of different products and the operation efficiency of the platform are further improved.

[0114] In an exemplary embodiment, the resource evaluation result obtaining module is further configured to: determine the resource interaction platform to which the target product belongs; based on the resource interaction platform, obtain object behavior data associated with the target product; according to the product attribute characteristics, the object behavior data, and the product market data, determine the resource interaction status data corresponding to the target product; based on the resource interaction platform, obtain the resource interaction record associated with the target product, and according to the resource interaction record and the object behavior data, determine the reward signal parameter corresponding to the target product; according to the resource interaction status data and the reward signal parameter, determine the pricing adjustment strategy corresponding to the target product.

[0115] In an exemplary embodiment, the product quality identification module is further configured to: extract features from the defect detection result to obtain the product defect features corresponding to the target product; obtain the product evaluation information corresponding to the target product, perform semantic analysis and feature extraction on the product evaluation information to obtain the product evaluation features corresponding to the target product; combine the product attribute features, the product evaluation features, and the product defect features to obtain the defect features to be identified corresponding to the target product; obtain the quality configuration information associated with the target product, and according to the quality configuration information, perform product quality identification on the defect features to be identified to obtain the quality identification result corresponding to the target product.

[0116] In an exemplary embodiment, the initial resource evaluation module is further configured to: perform data fusion on the defect detection result, the quality identification result, and the product market data to obtain the multi-modal product features corresponding to the target product; according to the trained resource evaluation model, perform product identification and resource evaluation on the multi-modal product features to obtain the initial resource evaluation result corresponding to the target product.

[0117] In an exemplary embodiment, a product resource evaluation device is further provided with a resource evaluation model training module, configured to: collect labeled sample data carrying defect labels and historical resource evaluation results, and unsupervised learning data carrying product evaluation information; according to the labeled sample data and the unsupervised learning data, perform model training on the initial resource evaluation model, and according to the reward signal parameter corresponding to the target product, perform fine-tuning on the initial resource evaluation model during the model training process, and when the model training end condition is met, obtain the trained resource evaluation model.

[0118] In an exemplary embodiment, the defect detection module is further configured to: perform defect detection on the product image features to obtain at least one defect corresponding to the product image features; according to the preset defect item configuration table, classify the at least one defect to obtain the defect category corresponding to each of the at least one defect, and determine the at least one defect and the defect category corresponding to each of the at least one defect as the defect detection result corresponding to the target product.

[0119] Each module in the above product resource evaluation device can be implemented in whole or in part by software, hardware, and their combination. Each of the above modules can be embedded in the processor of a computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0120] In an exemplary embodiment, a product resource evaluation system is provided, including a data collection module, a defect detection module, a color matching module, a price estimation module, a reinforcement learning module, a model training and optimization module, and a user interaction module, where:

[0121] 1. Data collection module: used to collect relevant market data from multiple channels, including the historical transaction prices of goods, user reviews, product images, and product descriptions, etc. Among them, the data sources can be internal transaction records, public data on third-party platforms, and other relevant data sources.

[0122] 2. Defect detection module: uses computer vision technology and large models to analyze product images and automatically detect defect items of the product. This module extracts features from product images through a pre-trained vision large model, and combines deep learning algorithms to locate and classify defects. Defect items include but are not limited to scratches, knocks, cracks, etc. Among them, the object detection module directly uses the pre-trained vision large model for preliminary detection, and then uses deep learning algorithms for fine detection and classification. The second step requires combining labeled data for model fine-tuning.

[0123] 3. Color matching module: according to the defect detection results and other features of the product (such as brand, model, service life, etc.), uses a large model to automatically match the color of the product. This module performs semantic analysis on product descriptions and user reviews through natural language processing technology, and combines color standards (such as nine-tenths new, eight-tenths new, etc.) for matching. Among them, by parsing and extracting the product description, specific defect item information is obtained, and then color matching is performed based on the defect item-color configuration table to obtain the color matching result.

[0124] 4. Price estimation module: combines the defect detection results, color matching results, and market data, and uses a price estimation large model for initial price estimation. This module integrates text, images, and structured data through multi-modal fusion technology and inputs them into the price estimation large model for training and prediction. Among them, the price estimation large model can learn complex non-linear relationships and the associations between multi-modal data, thereby improving the accuracy of price estimation. Specifically, feature extraction is performed through the price estimation large model, for example, text information is vectorized, image information is vectorized, and then they are input into a deep learning model for regression prediction to obtain the estimated price.

[0125] 5. Reinforcement Learning Module: Based on the initial price prediction, it uses reinforcement learning algorithms to dynamically adjust prices to optimize the platform's trading efficiency. This module models the platform's trading environment as a Markov Decision Process (MDP), where: 1) States specifically include product attribute features (referring to the information description of the commodity, including information such as the product's brand, model, service life, initial pricing, and release platform), product market data (which can specifically include the market supply and demand relationship for the target product), and object behavior data (whether the user clicks on the target product, the time when different users click on the target product, and the number of times different users click on the target product, as well as data such as the user's offer for the target product and the transaction price of the target product when a deal is reached). 2) Actions are specifically represented as price adjustment strategies in space (based on the premise that a deal can be reached after adjusting the product's pricing, the adjustment strategy is mainly based on two situations: sufficient exposure - no offer and sufficient offer - no deal. The reference price is lowered, and the intensity is adjusted based on the results given by reinforcement learning until a deal is made). 3) The reward function is specifically designed based on indicators such as the trading success rate, transaction time, and user satisfaction. Thus, by interacting with the resource interaction platform, the optimal price adjustment strategy of reinforcement learning is obtained to achieve the improvement of the platform's trading efficiency.

[0126] 6. Model Training and Optimization Module: Responsible for training and optimizing the large model and the reinforcement learning model. Through a large amount of labeled data and unsupervised learning data (labeled data includes: picture - defect item annotation, historical transaction prices, etc., and unsupervised data includes: user text descriptions), the large model is pre - trained and fine - tuned. At the same time, the model is optimized using the reward signal in reinforcement learning (the main signal is whether the lowered price results in a deal, the input is the sequence of historical price adjustments, and the label is "whether a deal is made", and the model is retrained), enabling it to better adapt to market changes and user needs during the pricing process.

[0127] 7. User Interaction Module: Provides a user interface to facilitate users to upload commodity information (such as pictures, descriptions, etc.) and display the pricing results. Users can provide feedback on the pricing results through this module, and the system further optimizes the model based on the user feedback.

[0128] In the above product resource evaluation system, by extracting the product image features of the target product to be detected, detecting defects in the product image features, obtaining the defect detection results corresponding to the target product, and acquiring the product attribute features corresponding to the target product, the product color identification of the target product is performed based on the product attribute features and the defect detection results, and the color identification results corresponding to the target product are obtained. Thus, a comprehensive and overall evaluation of the target product can be realized from the attributes of the target product itself, including defects and product attributes, etc., reducing the errors in the resource evaluation process. Further, by collecting the product market data associated with the target product, and based on the defect detection results, the color identification results, and the product market data, an initial resource evaluation of the target product is performed, and the initial resource evaluation results corresponding to the target product are obtained. Thus, a comprehensive evaluation of the target product can be carried out from multiple perspectives, improving the accuracy of the obtained initial resource evaluation results. After obtaining the initial resource evaluation results, by determining the pricing adjustment strategy corresponding to the target product, and further adjusting the initial resource evaluation results of the target product according to the pricing adjustment strategy, the dynamic adjustment of the resource evaluation results of the target product is realized, improving the accuracy of the obtained resource evaluation results corresponding to the target product, and further improving the resource interaction success rate of different products and the operation efficiency of the platform.

[0129] In an exemplary embodiment, a computer device is provided. The computer device can be a server or a terminal. Taking the computer device as a server as an example, its internal structure diagram can be as Figure 8 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as target products, product image features, defect detection results, product attribute features, color identification results, product market data, initial resource evaluation results, pricing adjustment strategies, and resource evaluation results. The input / output interface of the computer device is used for the processor to exchange information with external devices. The communication interface of the computer device is used to communicate with external terminals through a network connection. When the computer program is executed by the processor, a product resource evaluation method is implemented.

[0130] Those skilled in the art can understand, Figure 8The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0131] In one embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, the steps in the above method embodiments are implemented.

[0132] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0133] In one embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0135] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0136] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application. The above-described embodiments only represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of this application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application should be subject to the appended claims.

Claims

1. A product resource evaluation method, characterized in that, The method includes: Extracting the product image features of the target product to be detected, performing defect detection on the product image features, and obtaining a defect detection result corresponding to the target product; Obtaining the product attribute features corresponding to the target product, and based on the product attribute features and the defect detection result, performing product color recognition on the target product to obtain a color recognition result corresponding to the target product; Collecting product market data associated with the target product, and based on the defect detection result, the color recognition result, and the product market data, performing an initial resource evaluation on the target product to obtain an initial resource evaluation result corresponding to the target product; Determining a pricing adjustment strategy corresponding to the target product, and based on the pricing adjustment strategy, adjusting the initial resource evaluation result of the target product to obtain a resource evaluation result corresponding to the target product.

2. The method according to claim 1, characterized in that, The determining of the pricing adjustment strategy corresponding to the target product includes: Determining the resource interaction platform to which the target product belongs; Based on the resource interaction platform, obtaining object behavior data associated with the target product; Based on the product attribute features, the object behavior data, and the product market data, determining resource interaction status data corresponding to the target product; Based on the resource interaction platform, obtaining a resource interaction record associated with the target product, and based on the resource interaction record and the object behavior data, determining a reward signal parameter corresponding to the target product; Based on the resource interaction status data and the reward signal parameter, determining a pricing adjustment strategy corresponding to the target product.

3. The method according to claim 1, wherein The performing of product color recognition on the target product based on the product attribute features and the defect detection result to obtain a color recognition result corresponding to the target product includes: Performing feature extraction on the defect detection result to obtain product defect features corresponding to the target product; Obtaining product evaluation information corresponding to the target product, performing semantic analysis and feature extraction on the product evaluation information to obtain product evaluation features corresponding to the target product; Combining the product attribute features, the product evaluation features, and the product defect features to obtain defect features to be recognized corresponding to the target product; Obtaining color configuration information associated with the target product, and based on the color configuration information, performing product color recognition on the defect features to be recognized to obtain a color recognition result corresponding to the target product.

4. The method according to claim 1, wherein The performing of an initial resource evaluation on the target product based on the defect detection result, the color recognition result, and the product market data to obtain an initial resource evaluation result corresponding to the target product includes: Performing data fusion on the defect detection result, the color recognition result, and the product market data to obtain multi-modal product features corresponding to the target product; Based on the trained resource evaluation model, performing product recognition and resource evaluation on the multi-modal product features to obtain an initial resource evaluation result corresponding to the target product.

5. The method according to claim 4, wherein The training to obtain the resource evaluation model includes: Collect labeled sample data with defect annotations and historical resource evaluation results, as well as unsupervised learning data with product evaluation information; According to the labeled sample data and the unsupervised learning data, train an initial resource evaluation model, and according to the reward signal parameters corresponding to the target product, fine-tune the initial resource evaluation model during the model training process. When the model training end condition is met, obtain a trained resource evaluation model.

6. The method according to any one of claims 1 to 5, characterized in that The defect detection of the product image features to obtain a defect detection result corresponding to the target product includes: Perform defect detection on the product image features to obtain at least one defect corresponding to the product image features; According to a preset defect item configuration table, classify the at least one defect to obtain a defect category corresponding to each of the at least one defect, and determine the at least one defect and the defect category corresponding to each of the at least one defect as the defect detection result corresponding to the target product.

7. A product resource evaluation device, characterized in that, The device includes: A defect detection module for extracting product image features of a target product to be detected, performing defect detection on the product image features, and obtaining a defect detection result corresponding to the target product; A product color identification module for obtaining product attribute features corresponding to the target product, and performing product color identification on the target product according to the product attribute features and the defect detection result to obtain a color identification result corresponding to the target product; An initial resource evaluation module for collecting product market data associated with the target product, and performing an initial resource evaluation on the target product according to the defect detection result, the color identification result, and the product market data to obtain an initial resource evaluation result corresponding to the target product; A resource evaluation result obtaining module for determining a pricing adjustment strategy corresponding to the target product, and adjusting the initial resource evaluation result of the target product according to the pricing adjustment strategy to obtain a resource evaluation result corresponding to the target product.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images