CN115937603A - Commodity attribute detection method, device, storage medium and computer equipment - Google Patents

Abstract

The present application provides a product attribute detection method, device, storage medium and computer equipment. The method includes: obtaining multiple product pictures of the product to be detected; inputting each product picture into the posture detection model to obtain each The posture key point data corresponding to each commodity picture; according to the preset picture screening rules and the posture key point data corresponding to each commodity picture, filter out the first target picture; the posture key point data corresponding to each first target picture respectively Input into the binary classification model to obtain the classification results of each first target picture, each classification result is used to reflect whether the corresponding first target picture is suitable for attribute detection; according to the classification results of each first target picture to filter out second target pictures suitable for attribute detection; and detect predicted commodity attribute values of the commodity to be detected based on each second target picture. This application can improve the accuracy of attribute detection.

Description

Commodity attribute detection method, device, storage medium and computer equipment

technical field

The present application relates to the field of artificial intelligence, in particular to a commodity attribute detection method, device, storage medium and computer equipment.

Background technique

In the e-commerce platform, in order to facilitate users to quickly understand the product and to push the user personalizedly, it is necessary to determine the attribute value of each attribute of each product. Taking clothing products as an example, clothing products can include product attributes such as size attributes, length attributes, style attributes, and sleeve length attributes. Users can quickly understand the products through these product attributes. However, the existing commodity attribute detection methods have the problem of low accuracy.

Contents of the invention

The purpose of this application is to at least solve one of the above-mentioned technical defects, especially the technical defect of low detection accuracy in the prior art.

In the first aspect, the embodiment of the present application provides a method for detecting commodity attributes, the method comprising:

Obtain multiple product pictures of the product to be detected;

Obtain the pose detection model and the binary classification model;

Inputting each of the commodity pictures into the posture detection model respectively, so as to obtain the posture key point data corresponding to each of the commodity pictures;

According to the preset image screening rules and the posture key point data corresponding to each of the product images, the first target image is selected from each of the product images;

Input the posture key point data corresponding to each of the first target pictures into the binary classification model to obtain the classification results of each of the first target pictures, and each of the classification results is used to reflect the corresponding whether the first target image is suitable for attribute detection;

According to the classification results of each of the first target pictures, a second target picture suitable for attribute detection is selected from each of the first target pictures;

Detecting predicted commodity attribute values of the commodity to be detected based on each of the second target pictures.

In one of the embodiments, the step of detecting the predicted product attribute value of the product to be detected based on each of the second target pictures includes:

Determine the attribute to be detected, and obtain the single-label multi-classification model corresponding to the attribute to be detected;

Inputting each of the second target pictures into the single-label multi-classification model to obtain each detection attribute value output by the single-label multi-classification model and the confidence degree corresponding to each detection attribute value;

Based on each of the detected attribute values and the confidence degree corresponding to each of the detected attribute values, the predicted commodity attribute value is determined in each of the detected attribute values.

In one of the embodiments, the commodity attribute detection method further includes:

Acquiring the initial product attribute value of the product to be detected that is manually pre-labeled;

If the initial commodity attribute value is different from the predicted commodity attribute value, and the confidence corresponding to the predicted commodity attribute value is greater than a preset confidence threshold, attribute modification information is pushed to the information maintenance personnel.

In one of the embodiments, the single-label multi-classification model is trained using the Focal Loss function as a loss function.

In one of the embodiments, the step of selecting the first target picture from each of the product pictures according to the preset picture screening rules and the gesture key point data corresponding to each of the product pictures includes:

According to the posture key point data corresponding to each of the product pictures, it is judged whether each of the product pictures is a model display picture, and each of the model display pictures is used as the first target picture.

In one of the embodiments, the step of obtaining the binary classification model includes:

Obtain an initial linear regression model and a training data set, the training data set includes a plurality of training pictures and the manual classification results of each of the training pictures marked in advance;

Input each of the training pictures into the attitude detection model to obtain the attitude key point data corresponding to each of the training pictures;

Input the posture key point data corresponding to each of the training pictures into the initial linear regression model to obtain the training classification results corresponding to each of the training pictures, and according to each of the training classification results and the manual For classification results, the initial linear regression model is iteratively trained until the preset training completion condition is met and the binary classification model is obtained.

In the second aspect, the embodiment of the present application provides a product attribute detection device, the device includes:

Commodity picture acquisition module, used to obtain multiple commodity pictures of the commodity to be detected;

Two classification model acquisition modules, used to obtain attitude detection models and two classification models;

A posture detection module, configured to input each of the commodity pictures into the posture detection model, so as to obtain the posture key point data corresponding to each of the commodity pictures;

The first target picture screening module is used to filter out the first target picture from each of the commodity pictures according to the preset picture screening rules and the posture key point data corresponding to each of the commodity pictures;

A classification module, configured to input the posture key point data corresponding to each of the first target pictures into the binary classification model, so as to obtain a classification result of each of the first target pictures, and each of the classification results Used to reflect whether the corresponding first target picture is suitable for attribute detection;

The second target picture screening module is used to select a second target picture suitable for attribute detection from each of the first target pictures according to the classification results of each of the first target pictures;

An attribute value detection module, configured to determine the predicted commodity attribute value of the commodity to be detected according to each of the second target pictures.

In one of the embodiments, the attribute value detection module includes:

A single-label multi-classification model acquisition unit, configured to determine the attribute to be detected, and obtain a single-label multi-classification model corresponding to the attribute to be detected;

a detection unit, configured to input each of the second target pictures into the single-label multi-classification model, so as to obtain each detection attribute value output by the single-label multi-classification model and each detection attribute value corresponding to Confidence;

The attribute value determining unit is configured to determine the predicted commodity attribute value in each of the detected attribute values based on each of the detected attribute values and the confidence degree corresponding to each of the detected attribute values.

In a third aspect, the embodiment of the present application provides a storage medium, the storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, one or more processors Execute the steps of the commodity attribute detection method described in any one of the above embodiments.

In a fourth aspect, the embodiment of the present application provides a computer device, including: one or more processors, and a memory;

Computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the one or more processors, the steps of the commodity attribute detection method in any of the above-mentioned embodiments are executed.

In the commodity attribute detection method, device, storage medium and computer equipment of the present application, by introducing a posture detection model and a binary classification model, posture detection can be performed on each commodity picture, and based on the posture key point data output by the posture detection model and The binary classification model determines whether each product picture is suitable for attribute detection, so as to filter out the product pictures suitable for attribute detection from multiple product pictures, and perform attribute detection accordingly. In this way, the interference of inappropriate product pictures on attribute detection can be eliminated, thereby improving the accuracy of attribute detection.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present application. Those skilled in the art can also obtain other drawings based on these drawings without any creative effort.

Fig. 1 is one of the schematic flow charts of the product attribute detection method in an embodiment;

Fig. 2 is a schematic flow diagram of the step of obtaining a two-category model in one embodiment;

Fig. 3 is a schematic flow chart of the step of detecting the predicted product attribute value of the product to be detected based on each of the second target pictures in an embodiment;

Fig. 4 is the second schematic flow diagram of the product attribute detection method in an embodiment;

Fig. 5 is a schematic structural diagram of a product attribute detection device in an embodiment;

Fig. 6 is a schematic structural diagram of a computer device in an embodiment.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the application with reference to the drawings in the embodiments of the application. Apparently, the described embodiments are only some of the embodiments of the application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

As mentioned in the background art, the existing commodity attribute detection methods have the problem of low accuracy. The inventor found that the reason for this problem is that in the process of product attribute detection based on product pictures, since there are multiple product pictures for the same product, the display methods and display contents of each product picture are different, so Some product images are not suitable for product attribute detection. For example, the same clothing product may have a model display image, a partial display image, and a fabric display image, etc., and the model display image may include multiple product images in different postures, such as an upright display image, a sitting image display image, and the like. However, the existing technology cannot judge whether each product picture is suitable for product detection, but can only randomly select any product picture to complete the attribute detection, or use all product pictures to complete the attribute detection, resulting in the result of the attribute detection being different from the actual situation. The deviation is large, which leads to the problem of low accuracy.

In order to solve the above problems, this application provides a product attribute detection method, device, storage medium and computer equipment. By introducing a posture detection model and a binary classification model to process each product picture separately, it is possible to determine whether each product picture is suitable. It is used for attribute detection, so as to filter out suitable product images from each product image and perform attribute detection accordingly. In this way, the interference of inappropriate product pictures on attribute detection can be eliminated, thereby improving the accuracy of attribute detection.

In one embodiment, the present application provides a product attribute detection method. The following embodiments are described by taking the method applied to computer equipment as an example. It can be understood that computer equipment may be a device with data processing functions, but not It is limited to personal laptops, desktop computers, single servers or server clusters, etc. As shown in Figure 1, the method may include the following steps:

S102: Obtain multiple product pictures of the product to be detected.

Wherein, the commodity to be detected refers to the commodity whose attribute detection is to be performed, such as clothing commodity and the like. On the e-commerce platform, suppliers can upload multiple product pictures of the same product to display product details through multiple product pictures. In this step, the computer device may obtain multiple product images of the same product.

S104: Obtain a gesture detection model and a binary classification model.

Wherein, the pose detection model refers to a model for detecting the pose of a human body, which can perform pose detection on an input picture, and output pose key point data, so as to represent the human pose in the input picture through the pose key point data.

A binary classification model refers to a model used to detect whether an input image is suitable for attribute detection. Specifically, the binary classification model can receive an input picture and output a classification result of the input picture. The classification result can be used to reflect whether the input picture is suitable for attribute detection. For example, the classification result can be "suitable" or "inappropriate". suitable".

In one of the embodiments, as shown in Figure 2, the step of obtaining the binary classification model includes:

S202: Acquire an initial linear regression model and a training data set, where the training data set includes a plurality of training pictures and a pre-labeled manual classification result of each of the training pictures.

Wherein, the initial linear regression model refers to an untrained linear regression model. The multiple training pictures may be pictures of models in different poses and different shooting angles, such as upright pictures, sitting pictures, etc. Each training picture corresponds to a manual classification result, which refers to a result that is manually pre-labeled and used to reflect whether the training picture is suitable for attribute detection.

S204: Input each of the training pictures into the pose detection model to obtain pose key point data corresponding to each of the training pictures.

In this step, for each training picture, the computer device can input the training picture into the attitude detection model, so as to perform attitude detection on the training picture through the attitude detection model, and output the corresponding training picture used to represent the training picture. The attitude key point data of the human body posture in the human body posture, so that the result relationship data of the attitude key point data and the manual classification result can be obtained.

S206: Input the posture key point data corresponding to each of the training pictures into the initial linear regression model to obtain the training classification results corresponding to each of the training pictures, and according to each of the training classification results and the The manual classification result is used to iteratively train the initial linear regression model until the preset training completion conditions are met and the binary classification model is obtained.

In this step, the computer device can iteratively train the initial linear regression model by using each set of training data including posture key point data and manual classification results, until the loss performance of the initial linear regression model meets the preset training completion conditions. If the loss performance of the initial linear regression model meets the preset training completion conditions, it indicates that the initial linear regression model has been trained. The initial linear regression model is a binary classification model that can accurately detect whether the input image is suitable for attribute detection.

The present application obtains the binary classification model through the steps shown in S202 to S206, so that the classification accuracy of the binary classification model can be improved, so as to further improve the accuracy of commodity attribute detection.

S106: Input each of the product pictures into the pose detection model to obtain pose key point data corresponding to each of the product pictures.

In this step, for each commodity picture of the commodity to be detected, the computer device can input the commodity picture into the posture detection model, so as to perform posture detection on the commodity picture through the posture detection model, and output the corresponding Pose key point data representing the posture of the human body in the product image.

S108: According to the preset image screening rules and the posture key point data corresponding to each of the product images, filter out the first target image from each of the product images.

Among them, the image screening rule refers to the rule used to filter the image based on the posture key point data, and the specific content of the rule can be determined according to the actual situation, and this application does not make specific restrictions on this. In this step, the computer device can perform picture screening on each product picture based on the preset picture screening rules based on the gesture key point data corresponding to each product picture, and use each product picture that satisfies the picture screening rules as Each image of the first target.

In one of the embodiments, S108 may include: according to the posture key point data corresponding to each of the product pictures, respectively judge whether each of the product pictures is a model display picture, and use each of the model display pictures as the Describe the first target image.

Specifically, among the multiple product pictures of the product to be detected, there may be various types of pictures such as model display images, partial display images, and fabric display images. For product attributes that cannot be identified from partial display images and fabric display images, such as garment length and sleeve length, computer equipment can preliminarily screen out model display images from multiple product images, and exclude partial display images and fabric display images and other types of product images. In this way, on the one hand, the number of pictures processed by the binary classification model can be reduced, and the detection efficiency can be improved. On the other hand, the interference of the non-model display image on the binary classification model can also be eliminated, so that the classification result output by the binary classification model is more accurate, so as to further improve the detection accuracy.

In this application, for each product picture of the product to be detected, the computer device can judge whether there is a model in the product picture according to the pose key point data of the product picture output by the pose detection model, and then judge whether the product picture is a model If it is a display picture, then the product picture is used as the first target picture. In this way, the model display image can be selected from each commodity image as the first target image.

S110: Input the posture key point data corresponding to each of the first target pictures into the binary classification model to obtain the classification results of each of the first target pictures, and each of the classification results is used to reflect Whether the corresponding first target picture is suitable for attribute detection.

Specifically, for each first target picture, the computer device can input the posture key point data corresponding to the first target picture into the binary classification model, so as to obtain the classification result of the first target picture output by the binary classification model, Furthermore, it may be determined whether the first target picture is suitable for attribute detection based on the classification result of the first target picture.

This application first conducts posture detection on product pictures, and then performs binary classification based on the posture key point data obtained from posture detection to determine whether the product picture is suitable for attribute detection, thereby improving the accuracy of the binary classification results and improving product attributes. detection accuracy.

S112: According to the classification results of each of the first target pictures, select a second target picture suitable for attribute detection from each of the first target pictures.

After obtaining the classification results of each first target picture, the computer device can judge whether the first target picture is suitable for attribute detection based on the classification result of each first target picture, and accordingly screen out suitable The first target picture for attribute detection is used as the second target picture.

S114: Detect predicted commodity attribute values of the commodity to be detected based on each of the second target pictures.

In this step, the computer device can obtain the predicted commodity attribute of the commodity to be detected according to the second target picture suitable for attribute detection among the multiple commodity pictures. It can be understood that step S114 can be implemented in any manner in the present application. In one of the embodiments, as shown in FIG. 3, S114 may include the following steps:

S302: Determine the attribute to be detected, and acquire a single-label multi-classification model corresponding to the attribute to be detected.

Wherein, the attribute to be detected refers to the attribute of the commodity to be detected, which may be, but not limited to, the attribute of the commodity such as length of clothing, length of sleeves, and style of clothing. After determining the attribute to be detected, the computer device can obtain a single-label multi-classification model corresponding to the attribute to be detected, so as to perform attribute detection on the product to be detected by the single-label multi-classification model to obtain the attribute value of the attribute to be detected.

It can be understood that the present application can implement a single-label multi-classification model based on any neural network, and can also use any method to train the initial model to obtain a single-label multi-classification model. In one of the embodiments, during the training process of the initial model, the application can use the Focal Loss function as the loss function for training, so as to solve the problem of unbalanced training samples.

S304: Input each of the second target pictures into the single-label multi-classification model to obtain each detection attribute value output by the single-label multi-classification model and a confidence degree corresponding to each detection attribute value.

Wherein, the single-label multi-classification model may output at least one detection attribute value and a confidence degree corresponding to each detection attribute value according to the acquired input image. For example, the output data of the single-label multi-classification model may include a first attribute value, a confidence degree corresponding to the first attribute value, a second attribute value, and a confidence degree corresponding to the second attribute value.

The computer device can input each second target picture into the single-label multi-classification model, so as to perform attribute detection on each second target picture through the single-label multi-classification model, and output at least one detected attribute value and Confidence for each detected attribute value.

S306: Based on each of the detected attribute values and the confidence corresponding to each of the detected attribute values, determine the predicted commodity attribute value in each of the detected attribute values.

For example, the computer device may use the detection attribute value corresponding to the highest confidence level as the predicted commodity attribute value of the commodity to be detected.

The present application implements S114 through S302 to S306, so that the predicted product attribute value of the product to be detected can be automatically obtained through artificial intelligence, and the detection efficiency can be improved.

This application introduces a posture detection model and a binary classification model, so that posture detection can be performed on each product picture, and based on the posture key point data output by the posture detection model and the binary classification model, it is determined whether each commodity picture is suitable for Attribute detection, so as to filter out product images suitable for attribute detection from multiple product images, and perform attribute detection accordingly. In this way, the interference of inappropriate product pictures on attribute detection can be eliminated, thereby improving the accuracy of attribute detection.

In one embodiment, as shown in FIG. 4, the product attribute detection method of the present application further includes the following steps:

S402: Obtain the initial product attribute value of the product to be detected that is manually pre-labeled;

S404: If the initial commodity attribute value is different from the predicted commodity attribute value, and the confidence corresponding to the predicted commodity attribute value is greater than a preset confidence threshold, push attribute modification information to information maintenance personnel.

Specifically, when uploading product information, the supplier may manually mark the initial product attribute value of the product to be detected. The computer equipment can compare the manually marked initial commodity attribute value with the predicted commodity attribute value obtained by automatic identification, so as to determine whether the initial commodity attribute value needs to be corrected. If the initial product attribute value is different from the predicted product attribute value, and the confidence corresponding to the predicted product attribute value is greater than the preset confidence threshold, it indicates that the predicted product attribute value is relatively accurate, and the initial product attribute value has a higher probability of being different from the actual situation. If there is a deviation, in this case, the computer device can push attribute modification information to the information maintenance personnel to remind the information maintenance personnel to modify the commodity attribute value marked on the commodity information. In this way, product attributes can be checked, and attribute modification information can be pushed to information maintenance personnel when there is a high probability that product attributes are incorrect, thereby reducing the occurrence of product information errors as much as possible.

The product attribute detection device provided in the embodiment of the present application is described below, and the product attribute detection device described below and the product attribute detection method described above can be referred to in correspondence.

In one embodiment, the present application provides a product attribute detection device 500 . As shown in Figure 5, the device 500 includes a commodity image acquisition module 510, a binary classification model acquisition module 520, a posture detection module 530, a first target image screening module 540, a classification module 550, a second target image screening module 560 and an attribute value detection module 570 . in:

Commodity picture acquisition module 510, used to obtain multiple commodity pictures of the commodity to be detected;

Two classification model obtaining module 520, for obtaining posture detection model and two classification models;

The pose detection module 530 is configured to input each of the product pictures into the pose detection model, so as to obtain the pose key point data corresponding to each of the product pictures;

The first target picture screening module 540 is used to filter out the first target picture from each of the commodity pictures according to preset picture screening rules and posture key point data corresponding to each of the commodity pictures;

The classification module 550 is configured to respectively input the posture key point data corresponding to each of the first target pictures into the binary classification model, so as to obtain a classification result of each of the first target pictures, and each of the classification The result is used to reflect whether the corresponding first target picture is suitable for attribute detection;

The second target picture screening module 560 is configured to select a second target picture suitable for attribute detection from each of the first target pictures according to the classification results of each of the first target pictures;

The attribute value detection module 570 is configured to determine the predicted commodity attribute value of the commodity to be detected according to each of the second target pictures.

In one embodiment, the attribute value detection module 570 includes a single-label multi-classification model acquisition unit, a detection unit and an attribute value determination unit. Wherein, the single-label multi-classification model acquisition unit is used to determine the attribute to be detected, and obtain the single-label multi-classification model corresponding to the attribute to be detected; the detection unit is used to input each of the second target pictures into the unit In the label multi-classification model, to obtain each detection attribute value output by the single-label multi-classification model and the confidence degree corresponding to each detection attribute value; the attribute value determination unit is used to be based on each of the detection attribute values and each detection attribute value Confidence corresponding to each of the detected attribute values, and determine the predicted commodity attribute value in each of the detected attribute values.

In one embodiment, the product attribute detection device 500 further includes an attribute value acquisition module and an information push module. Wherein, the attribute value obtaining module is used to obtain the initial product attribute value of the product to be detected which is pre-marked manually. The information push module is used to push the attribute modification to the information maintenance personnel when the initial commodity attribute value is different from the predicted commodity attribute value, and the confidence corresponding to the predicted commodity attribute value is greater than the preset confidence threshold information.

In one embodiment, the single-label multi-classification model is trained using the Focal Loss function as a loss function.

In one embodiment, the first target image screening module 540 includes a model presentation image screening unit. The model display image screening unit is used to judge whether each of the product images is a model display image according to the posture key point data corresponding to each of the product images, and use each of the model display images as the first target image.

In one embodiment, the binary classification model acquisition module 520 includes an initial model acquisition unit, a training posture detection unit and a model training unit. Wherein, the initial model obtaining unit is used to obtain an initial linear regression model and a training data set, and the training data set includes a plurality of training pictures and a manual classification result of each training picture marked in advance. The training pose detection unit is used to input each of the training pictures into the pose detection model, so as to obtain the pose key point data corresponding to each of the training pictures. The model training unit is used to respectively input the posture key point data corresponding to each of the training pictures into the initial linear regression model, so as to obtain the training classification results corresponding to each of the training pictures, and according to each of the training classification As a result and the manual classification result, the initial linear regression model is iteratively trained until the preset training completion condition is met and the binary classification model is obtained.

In one embodiment, the present application also provides a storage medium, in which computer-readable instructions are stored, and when the computer-readable instructions are executed by one or more processors, one or more processing The device executes the steps of the product attribute detection method described in any of the above embodiments.

In an embodiment, the present application also provides a computer device. Computer-readable instructions are stored in the computer device, and when the computer-readable instructions are executed by one or more processors, one or more processors are made to perform the steps of the product attribute detection method in any of the above-mentioned embodiments.

Schematically, FIG. 6 is a schematic diagram of an internal structure of a computer device provided in an embodiment of the present application. In an example, the computer device may be a server. Referring to FIG. 6 , computer device 900 includes processing component 902 , which further includes one or more processors, and a memory resource represented by memory 901 for storing instructions executable by processing component 902 , such as application programs. The application program stored in memory 901 may include one or more modules each corresponding to a set of instructions. In addition, the processing component 902 is configured to execute instructions to execute the steps of the commodity attribute detection method described in any of the foregoing embodiments.

The computer device 900 may also include a power supply component 903 configured to perform power management of the computer device 900, a wired or wireless network interface 904 configured to connect the computer device 900 to a network, and an input output (I/O) interface 905 . The computer device 900 can operate based on an operating system stored in the memory 901, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™ or the like.

Those skilled in the art can understand that the internal structure of the computer equipment shown in this application is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation on the computer equipment on which the solution of this application is applied. The computer device may include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

Finally, it should also be noted that in this text, relational terms such as first and second etc. are only used to distinguish one entity or operation from another, and do not necessarily require or imply that these entities or operations, any such actual relationship or order exists. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element. As used herein, "a", "an", "the", "the" and "it" may also include plural forms unless the context clearly dictates otherwise. A plurality refers to at least two situations, such as 2, 3, 5 or 8 and so on. "And/or" includes any and all combinations of the associated listed items.

Each embodiment in this specification is described in a progressive manner. Each embodiment focuses on the difference from other embodiments. The various embodiments can be combined as needed, and the same and similar parts can be referred to each other. .

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (10)

1. A commodity attribute detection method, characterized in that the method comprises: Obtain multiple product pictures of the product to be detected; Obtain the pose detection model and the binary classification model; Inputting each of the commodity pictures into the posture detection model respectively, so as to obtain the posture key point data corresponding to each of the commodity pictures; According to the preset image screening rules and the posture key point data corresponding to each of the product images, the first target image is selected from each of the product images; Input the posture key point data corresponding to each of the first target pictures into the binary classification model to obtain the classification results of each of the first target pictures, and each of the classification results is used to reflect the corresponding whether the first target image is suitable for attribute detection; According to the classification results of each of the first target pictures, a second target picture suitable for attribute detection is selected from each of the first target pictures; Detecting predicted commodity attribute values of the commodity to be detected based on each of the second target pictures. 2. The commodity attribute detection method according to claim 1, wherein the step of detecting the predicted commodity attribute value of the commodity to be detected based on each of the second target pictures comprises: Determine the attribute to be detected, and obtain the single-label multi-classification model corresponding to the attribute to be detected; Inputting each of the second target pictures into the single-label multi-classification model to obtain each detection attribute value output by the single-label multi-classification model and the confidence degree corresponding to each detection attribute value; Based on each of the detected attribute values and the confidence degree corresponding to each of the detected attribute values, the predicted commodity attribute value is determined in each of the detected attribute values. 3. The commodity attribute detection method according to claim 2, wherein the method further comprises: Acquiring the initial product attribute value of the product to be detected that is manually pre-labeled; If the initial commodity attribute value is different from the predicted commodity attribute value, and the confidence corresponding to the predicted commodity attribute value is greater than a preset confidence threshold, attribute modification information is pushed to the information maintenance personnel. 4. The commodity attribute detection method according to claim 2, wherein the single-label multi-classification model is trained using FocalLoss function as a loss function. 5. The commodity attribute detection method according to claim 1, characterized in that, according to the preset image screening rules and the posture key point data corresponding to each of the commodity pictures, the product is screened in each of the commodity pictures The steps of producing the first target picture include: According to the posture key point data corresponding to each of the product pictures, it is judged whether each of the product pictures is a model display picture, and each of the model display pictures is used as the first target picture. 6. The commodity attribute detection method according to any one of claims 1 to 5, wherein the step of obtaining a binary classification model includes: Obtain an initial linear regression model and a training data set, the training data set includes a plurality of training pictures and the manual classification results of each of the training pictures marked in advance; Input each of the training pictures into the attitude detection model to obtain the attitude key point data corresponding to each of the training pictures; Input the posture key point data corresponding to each of the training pictures into the initial linear regression model to obtain the training classification results corresponding to each of the training pictures, and according to each of the training classification results and the manual For classification results, the initial linear regression model is iteratively trained until the preset training completion condition is met and the binary classification model is obtained. 7. A product attribute detection device, characterized in that the device comprises: Commodity picture acquisition module, used to obtain multiple commodity pictures of the commodity to be detected; Two classification model acquisition modules, used to obtain attitude detection models and two classification models; A posture detection module, configured to input each of the commodity pictures into the posture detection model, so as to obtain the posture key point data corresponding to each of the commodity pictures; The first target picture screening module is used to filter out the first target picture from each of the commodity pictures according to the preset picture screening rules and the posture key point data corresponding to each of the commodity pictures; A classification module, configured to input the posture key point data corresponding to each of the first target pictures into the binary classification model, so as to obtain a classification result of each of the first target pictures, and each of the classification results Used to reflect whether the corresponding first target picture is suitable for attribute detection; The second target picture screening module is used to select a second target picture suitable for attribute detection from each of the first target pictures according to the classification results of each of the first target pictures; An attribute value detection module, configured to determine the predicted commodity attribute value of the commodity to be detected according to each of the second target pictures. 8. The product attribute detection device according to claim 7, wherein the attribute value detection module comprises: A single-label multi-classification model acquisition unit, configured to determine the attribute to be detected, and obtain a single-label multi-classification model corresponding to the attribute to be detected; a detection unit, configured to input each of the second target pictures into the single-label multi-classification model, so as to obtain each detection attribute value output by the single-label multi-classification model and each detection attribute value corresponding to Confidence; The attribute value determining unit is configured to determine the predicted commodity attribute value in each of the detected attribute values based on each of the detected attribute values and the confidence degree corresponding to each of the detected attribute values. 9. A storage medium, characterized in that computer-readable instructions are stored in the storage medium, and when the computer-readable instructions are executed by one or more processors, one or more processors execute The steps of the product attribute detection method described in any one of 1 to 6. 10. A computer device, comprising: one or more processors, and a memory; Computer-readable instructions are stored in the memory, and when the computer-readable instructions are executed by the one or more processors, the steps of the product attribute detection method according to any one of claims 1 to 6 are executed.

Images