CN112837108A - Information processing method, device and electronic device - Google Patents

Abstract

Translated fromChinese

本申请实施例公开了信息处理方法、装置及电子设备，所述方法包括：获取至少一个用于对商品对象图片进行特征提取的神经网络模型；将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。通过本申请实施例，能够提高推荐信息的质量，降低资源浪费。

The embodiments of the present application disclose an information processing method, an apparatus, and an electronic device. The method includes: acquiring at least one neural network model for feature extraction of a picture of a commodity object; inputting the picture of a target commodity object into the neural network model , obtain the output result of the output layer of the neural network model, and the activation value of at least one middle layer; the activation value of the middle layer represents that the middle layer of the neural network model in the process of obtaining the output result from the commodity object The feature information extracted from the picture, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer; according to the output result of the output layer and the activation value of the intermediate layer, determine the feature value, and generate the feature vector of the commodity object image. Through the embodiments of the present application, the quality of the recommended information can be improved, and the waste of resources can be reduced.

Description

Translated fromChinese

信息处理方法、装置及电子设备Information processing method, device and electronic device

技术领域technical field

本申请涉及信息处理技术领域，特别是涉及信息处理方法、装置及电子设备。The present application relates to the technical field of information processing, and in particular, to information processing methods, devices, and electronic equipment.

背景技术Background technique

在商品对象信息服务系统中，经常会通过多种方式为用户提供商品对象的推荐信息。例如，在客户端首页中，可以通过“猜你喜欢”板块，以信息流等形式为用户提供推荐的商品对象信息。具体的推荐算法中的参考信息可以有多种，例如，系统可以根据消费者平时的浏览习惯与浏览记录、收藏的产品等信息来定位买家的风格喜好，据此推荐给消费者类似的商品对象，并且可以通过商品对象的图片进行展示。另外，系统还可以定位人群的消费层级，在推荐时可以根据商品对象的属性与价格等信息来确定合适的人群，达到用户预期的效果；再者，还可以综合商品对象的人气(包括点击率、收藏率、加购率等)等信息，向消费者用户推荐更优质的商品对象，等等。In the commodity object information service system, recommendation information of commodity objects is often provided to users in various ways. For example, on the homepage of the client, the "Guess what you like" section can be used to provide users with information about recommended product objects in the form of information flow and the like. There can be various kinds of reference information in the specific recommendation algorithm. For example, the system can locate buyers' style preferences according to consumers' usual browsing habits, browsing records, favorite products and other information, and recommend similar products to consumers accordingly. objects, and can be displayed through pictures of commodity objects. In addition, the system can also locate the consumption level of the crowd. When recommending, it can determine the appropriate crowd based on the attributes and price of the product object, so as to achieve the effect expected by the user; , collection rate, add-on rate, etc.) and other information, recommend higher-quality commodity objects to consumer users, and so on.

总之，现有技术中这种提供推荐信息的方式，主要根据商品对象的类目、价格、人气等信息，来为用户选择具体推荐的商品对象。但是，在具体实现时，这种推荐信息实际被用户点击或者购买的比例可能并不高，大部分的推荐信息可能会由于并不真正符合用户的喜好等原因而被忽略，以至于浪费了已经占用的计算、网络等各种资源。In a word, this method of providing recommended information in the prior art mainly selects a specific recommended commodity object for the user according to the category, price, popularity and other information of the commodity object. However, in the specific implementation, the actual percentage of such recommended information being clicked or purchased by users may not be high, and most of the recommended information may be ignored due to the fact that it does not really meet the user's preferences, so as to waste the already Occupied computing, network and other resources.

因此，如何进一步提高推荐信息的质量，降低资源浪费，成为需要本领域技术人员解决的技术问题。Therefore, how to further improve the quality of the recommended information and reduce the waste of resources has become a technical problem that needs to be solved by those skilled in the art.

发明内容SUMMARY OF THE INVENTION

本申请提供了信息处理方法、装置及电子设备，能够提高推荐信息的质量，降低资源浪费。The present application provides an information processing method, device and electronic device, which can improve the quality of recommended information and reduce waste of resources.

本申请提供了如下方案：This application provides the following solutions:

一种信息处理方法，包括：An information processing method, comprising:

获取至少一个用于对商品对象图片进行特征提取的神经网络模型；Acquire at least one neural network model for feature extraction on commodity object images;

将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Input the target commodity object image into the neural network model, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer represents the process of obtaining the output result The feature information extracted by the middle layer of the neural network model from the image of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。According to the output result of the output layer and the activation value of the intermediate layer, the feature values in multiple dimensions are determined, and the feature vector of the commodity object image is generated.

一种确定用户特征信息的方法，包括：A method for determining user characteristic information, comprising:

确定与用户的历史行为信息关联的多个商品对象的图片信息；Determine the picture information of multiple commodity objects associated with the user's historical behavior information;

获取至少一个用于对商品对象图片进行特征提取的神经网络模型；Acquire at least one neural network model for feature extraction on commodity object images;

将多个商品对象的图片分别输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Input the pictures of a plurality of commodity objects into the neural network model respectively, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer indicates that the output result is obtained after the The feature information extracted by the middle layer of the neural network model from the picture of the commodity object in the process of , or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

通过对所述多个商品对象的图片分别对应的输出层的输出结果以及所述中间层的激活值进行统计，确定多个维度上的特征值，以用于生成所述用户的特征向量。By counting the output results of the output layer corresponding to the pictures of the multiple commodity objects and the activation values of the intermediate layer respectively, the feature values in multiple dimensions are determined to be used to generate the feature vector of the user.

一种商品对象推荐方法，包括：A product object recommendation method, including:

确定商品对象的目标推荐用户，以及商品对象推荐信息的来源数据库；Determine the target recommended users of the product object, and the source database of the recommended information of the product object;

获取所述用户的特征向量信息，以及所述来源数据库中多个商品对象的特征向量信息，所述商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information of the user and the feature vector information of multiple commodity objects in the source database, where the feature vectors of the commodity objects include the feature values of the pictures of the commodity objects in multiple dimensions, and the feature values Including: after the picture of the commodity object is input into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer The value is the feature information extracted from the picture of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

将所述用户的特征向量信息以及所述多个商品对象的特征向量信息输入到推荐算法中，以预测将所述商品对象推荐给所述用户后获得的点击率信息。The feature vector information of the user and the feature vector information of the plurality of commodity objects are input into the recommendation algorithm to predict the click rate information obtained after the commodity objects are recommended to the user.

一种商品对象推荐方法，包括：A product object recommendation method, including:

确定待推荐的商品对象的图片信息；Determine the picture information of the commodity object to be recommended;

获取所述商品对象的特征向量信息，所述商品对象的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information of the commodity object, the feature vector of the commodity object includes the feature values of the picture in multiple dimensions, and the feature value includes: by inputting the picture of the commodity object into at least one user After the neural network model for feature extraction is performed on the image of the commodity object, the output result of the output layer, and the activation value of at least one intermediate layer; the activation value of the intermediate layer is changed from the neural network model in the process of obtaining the output result. Feature information extracted from the picture of the commodity object, or feature information obtained by weighted summation and nonlinear transformation of the neurons in the upper layer;

获取多个用户的特征向量信息；Obtain feature vector information of multiple users;

将所述商品对象的特征向量信息以及所述多个用户的特征向量信息输入到推荐算法中，以预测将所述待推荐的商品对象推荐给所述用户后获得的点击率信息。The feature vector information of the commodity object and the feature vector information of the multiple users are input into the recommendation algorithm to predict the click rate information obtained after the commodity object to be recommended is recommended to the user.

一种商品对象的提供方法，包括：A method for providing a commodity object, including:

在根据目标商品对象提供相似商品对象信息的过程中，获取所述目标商品对象的特征向量信息，以及来源数据库中多个商品对象的特征向量信息；其中，商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；In the process of providing similar commodity object information according to the target commodity object, the feature vector information of the target commodity object and the feature vector information of a plurality of commodity objects in the source database are obtained; wherein, the feature vector of the commodity object includes the The feature values of the picture in multiple dimensions, the feature values include: after the picture of the commodity object is input into at least one neural network model for feature extraction on the picture of the commodity object, the output result of the output layer, and The activation value of at least one middle layer; the activation value of the middle layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the weighting of the neurons in the upper layer Summation and feature information obtained by nonlinear transformation;

通过将所述目标商品对象的特征向量信息与来源数据库中多个商品对象的特征向量信息进行相似性比对，提供与所述指定商品对象相似度符合条件的商品对象信息。By comparing the feature vector information of the target product object with the feature vector information of multiple product objects in the source database, the product object information that meets the conditions of similarity with the specified product object is provided.

一种提供商品对象信息的方法，包括：A method of providing commodity object information, comprising:

接收对目标页面的访问请求，所述目标页面关联有多个资源位，用于展示多个商品对象的信息，其中，所述商品对象关联有多张不同的图片；receiving an access request to a target page, the target page is associated with a plurality of resource bits for displaying information of a plurality of commodity objects, wherein the commodity objects are associated with a plurality of different pictures;

获取所述多张不同的图片分别对应的特征向量信息，以及访问者用户的特征向量信息；其中，所述图片对应的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information corresponding to the multiple different pictures respectively, and the feature vector information of the visitor user; wherein, the feature vector corresponding to the picture includes the feature values of the picture in multiple dimensions, and the feature The value includes: the output result of the output layer and the activation value of at least one middle layer after inputting the picture into at least one neural network model for feature extraction of the commodity object picture; the activation value of the middle layer is The feature information extracted by the neural network model from the picture of the commodity object in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

将所述访问者用户的特征向量信息以及所述商品对象的多张图片对应的特征向量信息分别输入到预测算法中，以预测将所述商品对象的多张图片展示给所述用户后分别获得的点击率信息；The feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object are respectively input into the prediction algorithm, so as to predict that the multiple pictures of the commodity object are displayed to the user and obtained respectively. click-through rate information;

根据预测结果，将点击率符合条件的图片作为对应商品对象的代表图片展示在所述目标页面对应的资源位中。According to the prediction result, a picture whose click rate meets the conditions is displayed in the resource position corresponding to the target page as a representative picture of the corresponding commodity object.

一种信息推荐方法，包括：An information recommendation method including:

确定目标商品对象关联的图片；Determine the picture associated with the target product object;

通过对所述图片进行特征分析，确定所述图片中的商品对象对应的展示场景类别信息；Determine the display scene category information corresponding to the commodity object in the picture by analyzing the characteristics of the picture;

确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；determining a set of users, and acquiring user feature information corresponding to multiple users in the user set, where the user feature information includes category information of display scenarios that the user is interested in;

根据所述图片对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；Determine the target user according to the display scene category information corresponding to the picture and the display scene category information that the user is interested in;

将所述图片推荐到所述目标用户关联的客户端。The picture is recommended to the client associated with the target user.

一种信息推荐方法，包括：An information recommendation method including:

确定目标商品对象关联的视频；Determine the video associated with the target product object;

从所述视频中提取至少一帧图像，通过对所述图像进行特征分析，确定所述视频中的商品对象对应的展示场景类别信息；Extract at least one frame of image from the video, and determine the display scene category information corresponding to the commodity object in the video by analyzing the characteristics of the image;

确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；determining a set of users, and acquiring user feature information corresponding to multiple users in the user set, where the user feature information includes category information of display scenarios that the user is interested in;

根据所述视频对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；Determine the target user according to the display scene category information corresponding to the video and the display scene category information that the user is interested in;

将所述视频推荐到所述目标用户关联的客户端。The video is recommended to the client associated with the target user.

一种信息处理装置，包括：An information processing device, comprising:

模型获取单元，用于获取至少一个用于对商品对象图片进行特征提取的神经网络模型；a model obtaining unit, used for obtaining at least one neural network model for feature extraction on the image of the commodity object;

分析单元，用于将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；The analysis unit is used for inputting the target commodity object picture into the neural network model, and obtaining the output result of the output layer of the neural network model and the activation value of at least one middle layer; the activation value of the middle layer is characterized in obtaining In the process of outputting the result, the feature information extracted by the middle layer of the neural network model from the picture of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

图片特征向量生成单元，用于根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。The image feature vector generating unit is configured to determine feature values in multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer, and generate the feature vector of the commodity object image.

一种确定用户特征信息的装置，包括：A device for determining user feature information, comprising:

图片信息确定单元，用于确定与用户的历史行为信息关联的多个商品对象的图片信息；A picture information determining unit, configured to determine picture information of multiple commodity objects associated with the user's historical behavior information;

模型获取单元，用于获取至少一个用于对商品对象图片进行特征提取的神经网络模型；a model obtaining unit, used for obtaining at least one neural network model for feature extraction on the image of the commodity object;

分析单元，用于将多个商品对象的图片分别输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；an analysis unit, configured to input the pictures of a plurality of commodity objects into the neural network model respectively, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer Characterize the feature information extracted from the image of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

用户特征向量生成单元，用于通过对所述多个商品对象的图片分别对应的输出层的输出结果以及所述中间层的激活值进行统计，确定多个维度上的特征值，以用于生成所述用户的特征向量。A user feature vector generating unit, configured to determine feature values in multiple dimensions by performing statistics on the output results of the output layers corresponding to the pictures of the multiple commodity objects and the activation values of the intermediate layers, for generating The feature vector of the user.

一种商品对象信息推荐装置，包括：A product object information recommendation device, comprising:

信息确定单元，用于确定商品对象的目标推荐用户，以及商品对象推荐信息的来源数据库；an information determination unit, used for determining the target recommending user of the commodity object and the source database of the commodity object recommendation information;

特征向量获取单元，用于获取所述用户的特征向量信息，以及所述来源数据库中多个商品对象的特征向量信息，所述商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；A feature vector obtaining unit is used to obtain the feature vector information of the user and the feature vector information of a plurality of commodity objects in the source database, where the feature vectors of the commodity objects include pictures of the commodity objects in multiple dimensions. feature value, the feature value includes: after inputting the picture of the commodity object into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer and the activation value of at least one intermediate layer ; The activation value of the middle layer is the feature information extracted from the picture of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the weighted summation of the neurons in the upper layer and the Feature information obtained by linear transformation;

预测单元，用于将所述用户的特征向量信息以及所述多个商品对象的特征向量信息输入到推荐算法中，以预测将所述商品对象推荐给所述用户后获得的点击率信息。The prediction unit is configured to input the feature vector information of the user and the feature vector information of the plurality of commodity objects into the recommendation algorithm, so as to predict the click-through rate information obtained after recommending the commodity objects to the user.

一种商品对象推荐装置，包括：A product object recommendation device, comprising:

图片信息确定单元，用于确定待推荐的商品对象的图片信息；The picture information determination unit is used to determine the picture information of the commodity object to be recommended;

商品特征向量获取单元，用于获取所述商品对象的特征向量信息，所述商品对象的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；A product feature vector obtaining unit, configured to obtain feature vector information of the product object, where the feature vector of the product object includes feature values of the picture in multiple dimensions, and the feature values include: After the image of the object is input into at least one neural network model for feature extraction of the image of the commodity object, the output result of the output layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the neural network model The feature information extracted from the picture of the commodity object in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

用户特征向量获取单元，哟关于获取多个用户的特征向量信息；User feature vector acquisition unit, about acquiring feature vector information of multiple users;

预测单元，用于将所述商品对象的特征向量信息以及所述多个用户的特征向量信息输入到推荐算法中，以预测将所述待推荐的商品对象推荐给所述用户后获得的点击率信息。a prediction unit, configured to input the feature vector information of the commodity object and the feature vector information of the multiple users into the recommendation algorithm, so as to predict the click-through rate obtained after recommending the commodity object to be recommended to the user information.

一种商品对象的提供装置，包括：A device for providing commodity objects, comprising:

特征向量获取单元，用于在根据目标商品对象提供相似商品对象信息的过程中，获取所述目标商品对象的特征向量信息，以及来源数据库中多个商品对象的特征向量信息；其中，商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；The feature vector obtaining unit is used to obtain the feature vector information of the target product object and the feature vector information of multiple product objects in the source database in the process of providing similar product object information according to the target product object; The feature vector includes feature values of the picture of the commodity object in multiple dimensions, and the feature value includes: after inputting the picture of the commodity object into at least one neural network model for feature extraction of the picture of the commodity object, The output result of the output layer, and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the The feature information obtained by the weighted summation and nonlinear transformation of the neurons in the previous layer;

相似商品对象信息提供单元，用于通过将所述目标商品对象的特征向量信息与来源数据库中多个商品对象的特征向量信息进行相似性比对，提供与所述指定商品对象相似度符合条件的商品对象信息。A similar commodity object information providing unit is used to compare the feature vector information of the target commodity object with the feature vector information of a plurality of commodity objects in the source database, and provide the similarity with the specified commodity object. Product object information.

一种提供商品对象信息的装置，包括：A device for providing commodity object information, comprising:

访问请求接收单元，用于接收对目标页面的访问请求，所述目标页面关联有多个资源位，用于展示多个商品对象的信息，其中，所述商品对象关联有多张不同的图片；an access request receiving unit, configured to receive an access request to a target page, the target page is associated with a plurality of resource bits for displaying information of a plurality of commodity objects, wherein the commodity objects are associated with a plurality of different pictures;

特征向量获取单元，用于获取所述多张不同的图片分别对应的特征向量信息，以及访问者用户的特征向量信息；其中，所述图片对应的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；A feature vector obtaining unit, configured to obtain the feature vector information corresponding to the multiple different pictures respectively, and the feature vector information of the visitor user; wherein, the feature vectors corresponding to the pictures include the pictures in multiple dimensions The feature value of , the feature value includes: the output result of the output layer and the activation value of at least one intermediate layer after inputting the picture into at least one neural network model for feature extraction of commodity object pictures; The activation value of the middle layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the feature obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer. information;

预测单元，用于将所述访问者用户的特征向量信息以及所述商品对象的多张图片对应的特征向量信息分别输入到预测算法中，以预测将所述商品对象的多张图片展示给所述用户后分别获得的点击率信息；The prediction unit is used to input the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into the prediction algorithm respectively, so as to predict that the multiple pictures of the commodity object will be displayed to the The click-through rate information obtained after the user is described;

代表图片确定单元，用于根据预测结果，将点击率符合条件的图片作为对应商品对象的代表图片展示在所述目标页面对应的资源位中。The representative picture determining unit is configured to display, according to the prediction result, the picture whose click rate meets the conditions as the representative picture corresponding to the commodity object in the resource position corresponding to the target page.

一种信息推荐装置，包括：An information recommendation device, comprising:

图片确定单元，用于确定目标商品对象关联的图片；The picture determination unit is used to determine the picture associated with the target commodity object;

展示场景确定单元，用于通过对所述图片进行特征分析，确定所述图片中的商品对象对应的展示场景类别信息；a display scene determination unit, configured to determine the display scene category information corresponding to the commodity object in the picture by analyzing the characteristics of the picture;

用户特征信息获取单元，用于确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；a user feature information acquisition unit, configured to determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

目标用户确定单元，用于根据所述图片对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；a target user determination unit, configured to determine a target user according to the display scene category information corresponding to the picture and the display scene category information that the user is interested in;

图片推荐单元，用于将所述图片推荐到所述目标用户关联的客户端。A picture recommendation unit, configured to recommend the picture to the client associated with the target user.

一种信息推荐装置，包括：An information recommendation device, comprising:

视频确定单元，用于确定目标商品对象关联的视频；a video determination unit, used to determine the video associated with the target commodity object;

展示场景确定单元，用于从所述视频中提取至少一帧图像，通过对所述图像进行特征分析，确定所述视频中的商品对象对应的展示场景类别信息；a display scene determination unit, configured to extract at least one frame of image from the video, and determine the display scene category information corresponding to the commodity object in the video by performing feature analysis on the image;

用户特征信息获取单元，用于确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；a user feature information acquisition unit, configured to determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

目标用户确定单元，用于根据所述视频对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；a target user determining unit, configured to determine a target user according to the category information of the presentation scene corresponding to the video and the category information of the presentation scene that the user is interested in;

视频推荐单元，用于将所述视频推荐到所述目标用户关联的客户端。A video recommendation unit, configured to recommend the video to the client associated with the target user.

一种电子设备，包括：An electronic device comprising:

一个或多个处理器；以及one or more processors; and

与所述一个或多个处理器关联的存储器，所述存储器用于存储程序指令,所述程序指令在被所述一个或多个处理器读取执行时,执行如上各项所述的方法。memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform the method as described in each of the preceding items.

根据本申请提供的具体实施例，本申请公开了以下技术效果：According to the specific embodiments provided by the application, the application discloses the following technical effects:

通过本申请实施例，能够通过神经网络模型来对商品对象图片进行特征提取，并且不仅可以将模型输出层的输出结果作为图片特征向量中具体维度上的特征值，还可以将模型中间层的激活值作为特征向量中更多维度上的特征值。这样，可以为商品对象图片生成高维度的特征向量，其中不仅可以包括人类可以理解、定义的维度上的特征值(输出层的输出结果)，还可以包括人类可能无法理解或者准确定义的维度上的特征值(中间层的激活值)，从而使得这种特征向量能够更完整的表达图片的特征。进而在商品对象信息推荐等场景中，可以获得更准确的推荐结果，提高推荐信息的点击率，降低资源浪费。Through the embodiments of the present application, the neural network model can be used to perform feature extraction on the image of the commodity object, and not only the output result of the model output layer can be used as the feature value of the specific dimension in the image feature vector, but also the activation of the middle layer of the model can be used. values as eigenvalues on more dimensions in the eigenvector. In this way, a high-dimensional feature vector can be generated for the image of the commodity object, which can include not only the eigenvalues (the output result of the output layer) on the dimensions that humans can understand and define, but also the dimensions that humans may not understand or accurately define. The feature value (the activation value of the intermediate layer), so that this feature vector can more completely express the characteristics of the picture. Furthermore, in scenarios such as product object information recommendation, more accurate recommendation results can be obtained, the click-through rate of the recommendation information can be improved, and the waste of resources can be reduced.

当然，实施本申请的任一产品并不一定需要同时达到以上所述的所有优点。Of course, implementing any product of the present application does not necessarily need to achieve all of the advantages described above at the same time.

附图说明Description of drawings

为了更清楚地说明本申请实施例或现有技术中的技术方案，下面将对实施例中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the accompanying drawings required in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some of the present application. In the embodiments, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without any creative effort.

图1是本申请实施例提供的系统架构的示意图；1 is a schematic diagram of a system architecture provided by an embodiment of the present application;

图2是本申请实施例提供的第一方法的流程图；2 is a flowchart of a first method provided by an embodiment of the present application;

图3是本申请实施例提供的神经网络结构的示意图；3 is a schematic diagram of a neural network structure provided by an embodiment of the present application;

图4是本申请实施例提供的第二方法的流程图；4 is a flowchart of a second method provided by an embodiment of the present application;

图5是本申请实施例提供的第三方法的流程图；5 is a flowchart of a third method provided by an embodiment of the present application;

图6是本申请实施例提供的第四方法的流程图；6 is a flowchart of a fourth method provided by an embodiment of the present application;

图7是本申请实施例提供的第五方法的流程图；7 is a flowchart of a fifth method provided by an embodiment of the present application;

图8是本申请实施例提供的第六方法的流程图；8 is a flowchart of a sixth method provided by an embodiment of the present application;

图9是本申请实施例提供的第七方法的流程图；9 is a flowchart of a seventh method provided by an embodiment of the present application;

图10是本申请实施例提供的第八方法的流程图；10 is a flowchart of an eighth method provided by an embodiment of the present application;

图11是本申请实施例提供的第一装置的示意图；11 is a schematic diagram of a first device provided by an embodiment of the present application;

图12是本申请实施例提供的第二装置的示意图；12 is a schematic diagram of a second device provided by an embodiment of the present application;

图13是本申请实施例提供的第三装置的示意图；13 is a schematic diagram of a third device provided by an embodiment of the present application;

图14是本申请实施例提供的第四装置的示意图；14 is a schematic diagram of a fourth device provided by an embodiment of the present application;

图15是本申请实施例提供的第五装置的示意图；15 is a schematic diagram of a fifth device provided by an embodiment of the present application;

图16是本申请实施例提供的第六装置的示意图；16 is a schematic diagram of a sixth device provided by an embodiment of the present application;

图17是本申请实施例提供的第七装置的示意图；17 is a schematic diagram of a seventh device provided by an embodiment of the present application;

图18是本申请实施例提供的第八装置的示意图；18 is a schematic diagram of an eighth device provided by an embodiment of the present application;

图19是本申请实施例提供的电子设备的示意图。FIG. 19 is a schematic diagram of an electronic device provided by an embodiment of the present application.

具体实施方式Detailed ways

下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员所获得的所有其他实施例，都属于本申请保护的范围。The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art fall within the protection scope of this application.

首先需要说明的是，本申请发明人在实现本申请的过程中发现，商品对象的图片通常会具有一些特征。例如，图片中会有各自的主色调信息；另外，对于服装等类目的商品对象，拍摄场景也可能会有所不同，有的可能是在室外进行的“街拍”，有的可能是在室内摄影棚内进行的“棚拍”，有的有人物模特对使用状态进行展示，有的是在没有模特情况下进行的细节展示，另外，在有人物模特展示情况下，人物模特的展示范围、人体姿态、表情等也可能会不同。不同的特征向用户传达的调性不同，而不同的用户对各种不同特征的感兴趣程度也可能会有所不同。因此，在向用户提供一些推荐的商品对象信息时，被推荐的商品对象是否能够获得用户的点击等操作，除了与商品对象的类目、品牌等基础信息相关，与推荐出的商品对象的图片是否符合用户的兴趣点也是密切相关的。例如，某用户对于风衣这种商品对象，比较喜欢卡其色，并且在浏览商品对象信息的过程中，被该用户点击的商品对象的图片大部分具有以下特点：图片是在室内摄影棚中拍摄且由人物模特展示，人物模特为全身，等等。如果能够有效的利用上述信息进行商品对象信息的推荐，则推荐出的信息获得用户点击率的概率会比较高。First of all, it should be noted that the inventor of the present application found in the process of realizing the present application that the pictures of commodity objects usually have some features. For example, there will be their own main color information in the picture; in addition, for commodity objects such as clothing, the shooting scene may also be different, some may be "street photography" carried out outdoors, and some may be in the In the "studio shooting" in the indoor studio, some models are used to display the use status, and some are detailed displays without models. In addition, when there are models displayed, the display range of the models, the human body. Postures, expressions, etc. may also vary. Different features convey different tones to users, and different users may have different degrees of interest in various features. Therefore, when providing some recommended commodity object information to the user, whether the recommended commodity object can obtain the user's click and other operations, in addition to the basic information such as the category and brand of the commodity object, is related to the picture of the recommended commodity object. Whether it matches the user's point of interest is also closely related. For example, a user prefers khaki for a commodity object such as a windbreaker, and in the process of browsing the commodity object information, most of the pictures of the commodity object clicked by the user have the following characteristics: the pictures are taken in an indoor studio and Shown by mannequins, mannequins are full body, etc. If the above-mentioned information can be effectively used to recommend commodity object information, the probability of obtaining a user's click-through rate for the recommended information will be relatively high.

因此，具体实现时，可以通过一些图像处理算法，对商品对象图片进行特征提取。例如，包括图片的拍摄场景、有无人物模特或道具，甚至人物模特的展示范围、人体姿态、表情等，都可以通过深度学习等方式进行模型的训练，然后，通过训练好的模型实现对图片中对应特征的提取。这样，可以生成商品对象图片的特征向量，然后在具体的推荐算法中，可以通过商品对象图片的特征向量与用户特征向量等的运算，确定出具体的推荐方案。通过这种特征向量对商品对象图片特征进行描述，并输入到推荐算法中进行运算的方式，能够从一定程度上提高推荐信息的点击率。Therefore, in the specific implementation, some image processing algorithms can be used to extract features from the image of the commodity object. For example, including the shooting scene of the picture, the presence or absence of character models or props, and even the display range of the character models, human posture, expressions, etc., the model can be trained through deep learning and other methods, and then the trained model can be used to realize the picture. Extraction of corresponding features. In this way, the feature vector of the commodity object picture can be generated, and then in the specific recommendation algorithm, the specific recommendation scheme can be determined through the operation of the feature vector of the commodity object picture and the user feature vector. The method of describing the image features of the product object through this feature vector and inputting it into the recommendation algorithm for operation can improve the click-through rate of the recommendation information to a certain extent.

但是，通过上述方式进行图片特征提取时，能够获得的特征数量通常比较有限，相应的，生成的商品对象的特征向量的维度通常在十几或者几十数量级。并且，具体的特征通常都是一些比较形象的特征，包括前文所述的拍摄场景、有无人物模特或道具，甚至人物模特的展示范围、人体姿态、表情等，另外还可以包括图片的主色调、亮度，等等。这些形象的特征，通常都是按照人类的理解进行定义的，但是，实际上一张图片中通常还包含着更多的特征，这些特征可能是比较抽象的特征，甚至可能在人类所能理解、发现或者定义的范围之外。如果将这些抽象的特征也加入到商品对象图片的特征向量中，生成更高维度的特征向量，则可以更完整更丰富地表达图片的特征。进而，在将这种高维的特征向量输入到具体的推荐算法中进行运算时，有利于获得更精细、更准确的推荐信息，为推荐信息获得更好的点击率。However, when the image feature extraction is performed in the above manner, the number of features that can be obtained is usually limited, and accordingly, the dimension of the feature vector of the generated commodity object is usually in the order of ten or dozens. In addition, the specific features are usually some relatively vivid features, including the shooting scene mentioned above, the presence or absence of character models or props, and even the display range of the character models, human posture, expressions, etc., and can also include the main color of the picture. , brightness, etc. The features of these images are usually defined according to human understanding. However, in fact, a picture usually contains more features. These features may be relatively abstract features, and may even be within human comprehension, outside the scope of discovery or definition. If these abstract features are also added to the feature vector of the commodity object image to generate a higher-dimensional feature vector, the features of the image can be expressed more completely and abundantly. Furthermore, when such a high-dimensional feature vector is input into a specific recommendation algorithm for operation, it is beneficial to obtain finer and more accurate recommendation information, and to obtain a better click-through rate for the recommendation information.

为达到上述目的，本申请实施例可以通过神经网络模型来进行商品对象图片特征的提取，由于神经网络通常会包括多层，每层中可以包括多个神经元(其中，输入层与输出层的神经元数量往往是固定的，中间层则可以自由指定，例如：对于一个三层的神经网络而言，输入层有3个输入神经元，输出层有2个神经元，中间层可以有4个神经元，也可以有5个神经元，等等)。其中，输出层具体输出的结果，通常是根据训练过程中指定的训练方向来确定的。例如，某神经网络用于识别某图片的拍摄场景，则在将某具体的图片输入到该神经网络后，输出层的输出结果为该图片对应的拍摄场景是否为街拍或者棚拍，或者各种的概率分别是多少。但同时，神经网络还具有一些中间层，这些中间层中的每个神经元的激活值可以是从图片中抽取的各种特征，或者，由上一层神经元，经过加权求和与非线性变换而得到的特征，这些特征可能是人类所无法理解或者定义的抽象特征，但是，由于这些特征能够影响到输出层的输出结果，因此，也是有意义的。为此，在本申请实施例中，在利用神经网络对商品对象图片进行特征提取的过程中，除了可以根据输出层的输出结果来确定商品对象在其中一个或者多个维度上的特征值，另外，还可以获取其中部分或者全部中间层的输出值(通常称为激活值)，将这种中间层的激活值也作为商品对象在更多维度上的特征值。这样，实际为商品对象生成的特征向量中，既可以包括人类能够理解的比较形象化的一些维度上的特征值，还可以包括一些比较抽象的特征值，换言之，可以将抽象特征作为形象特征的补充，从而为商品对象获得高维度的特征向量。进而可以将这种高维度的特征向量输入到具体的推荐算法中，用于获得更精细化、更准确的推荐结果。In order to achieve the above purpose, the embodiment of the present application can use a neural network model to extract image features of commodity objects. Since a neural network usually includes multiple layers, each layer may include multiple neurons (wherein The number of neurons is often fixed, and the middle layer can be freely specified. For example, for a three-layer neural network, the input layer has 3 input neurons, the output layer has 2 neurons, and the middle layer can have 4 neurons. neuron, can also have 5 neurons, etc.). Among them, the specific output result of the output layer is usually determined according to the training direction specified in the training process. For example, if a neural network is used to identify the shooting scene of a certain picture, after a specific picture is input into the neural network, the output result of the output layer is whether the shooting scene corresponding to the picture is street shooting or studio shooting, or whether each What are the probabilities of each species. But at the same time, the neural network also has some intermediate layers, the activation value of each neuron in these intermediate layers can be various features extracted from the picture, or, by the neurons in the previous layer, after weighted summation and nonlinear The features obtained by transformation may be abstract features that humans cannot understand or define. However, because these features can affect the output results of the output layer, they are also meaningful. For this reason, in the embodiment of the present application, in the process of using neural network to perform feature extraction on the image of the commodity object, in addition to determining the feature value of the commodity object in one or more dimensions according to the output result of the output layer, in addition to , the output values of some or all of the intermediate layers (usually called activation values) can also be obtained, and the activation values of such intermediate layers are also used as the feature values of the commodity object in more dimensions. In this way, the feature vector actually generated for the commodity object can include not only the eigenvalues in some dimensions that can be understood by humans, but also some more abstract eigenvalues. Supplement to obtain a high-dimensional feature vector for the commodity object. Then, the high-dimensional feature vector can be input into a specific recommendation algorithm to obtain a more refined and accurate recommendation result.

具体实现时，从系统架构角度而言，参见图1，本申请实施例主要可以涉及商品对象信息服务系统的服务端以及客户端。其中，服务端可以生成具体的推荐信息，然后发送给对应用户的客户端，使得用户可以通过客户端浏览到具体的推荐信息。具体的，服务端中还可以提供特征提取模块，该模块中可以保存多个神经网络模型，分别可以用于从不同的角度对商品对象图片中包含的特征进行提取。例如，神经网络A用于识别图片中的商品对象主体区域以及背景区域，神经网络B用于识别图片的拍摄场景类别，等等。具体在需要提取某个商品对象的图片的特征信息时，可以将同一图片分别输入到多个神经网络中进行特征提取。另外，特征提取模块中还可以包括用于特征向量生成模块，该模块可以从上述多个神经网络中分别获得最终输出层的输出结果，以及一个或多个中间层的激活值，进而可以将这些输出结果以及激活值作为多个维度上的特征值，生成具体图片的特征向量。该特征向量可以输入到具体的推荐算法模块，以用于获得具体的推荐结果，并提供给客户端。当然，在实际应用中，通过本申请实施例提供的方案生成的关于商品对象的高维特征向量也可以应用到推荐系统之外的其他场景中，例如，向页面中进行商品对象信息投放时，可以根据上述特征向量信息，来确定具体使用哪个图片作为代表图片时能够获得更高的用户点击率，等等。During specific implementation, from the perspective of system architecture, referring to FIG. 1 , the embodiment of the present application may mainly involve the server and the client of the commodity object information service system. The server can generate specific recommendation information, and then send it to the client of the corresponding user, so that the user can browse the specific recommendation information through the client. Specifically, a feature extraction module may also be provided in the server, and a plurality of neural network models may be stored in the module, which may be respectively used to extract the features contained in the image of the commodity object from different angles. For example, the neural network A is used to identify the main area and background area of the commodity object in the picture, and the neural network B is used to identify the shooting scene category of the picture, and so on. Specifically, when the feature information of a picture of a certain commodity object needs to be extracted, the same picture can be input into multiple neural networks respectively for feature extraction. In addition, the feature extraction module can also include a feature vector generation module, which can obtain the output results of the final output layer and the activation values of one or more intermediate layers from the above-mentioned multiple neural networks, and then can use these The output results and activation values are used as feature values in multiple dimensions to generate feature vectors for specific images. The feature vector can be input into a specific recommendation algorithm module for obtaining specific recommendation results and provided to the client. Of course, in practical applications, the high-dimensional feature vectors of commodity objects generated by the solutions provided in the embodiments of the present application can also be applied to other scenarios outside the recommendation system. For example, when commodity object information is placed on a page, The above-mentioned feature vector information can be used to determine which picture can be used as the representative picture to obtain a higher user click rate, and so on.

下面对本申请实施例提供的具体实现方案进行详细介绍。The specific implementation solutions provided by the embodiments of the present application are described in detail below.

实施例一Example 1

首先，该实施例一提供了一种信息处理的方法，参见图2，该方法具体可以包括：First, the first embodiment provides an information processing method, see FIG. 2 , the method may specifically include:

S201：获取至少一个用于对商品对象图片进行特征提取的神经网络模型；S201: Acquire at least one neural network model for feature extraction on commodity object pictures;

在本申请实施例中，首先可以获得一个获得多个神经网络模型，这种模型可以是具有对商品对象图片进行特征提取功能的模型。其中，所述神经网络模型根据不同的特征识别对象具有对应的模型类别。之所以会存在多个模型的情况是因为，对于同一个图片而言，需要提取的特征可能是多方面的，而这些特征之间可能并不是完全互斥的关系，因此，难以通过同一个模型提取出这些特征。为此，可以通过多个不同的神经网络模型来从多种不同角度实现对商品对象图片的特征提取。例如，具体可以包括用于区分商品对象主题所在区域以及背景区域的模型，还可以包括用于识别拍摄场景是属于街拍还是棚拍的模型，还可以包括用于识别人物模特的展示范围的模型，等等。总之，对于同一张需要进行特征提取的图片而言，可以分别输入到多个不同的神经网络模型中，以从不同角度进行特征提取。In the embodiment of the present application, firstly, a plurality of neural network models can be obtained, and this model may be a model with the function of extracting features from pictures of commodity objects. Wherein, the neural network model identifies objects with corresponding model categories according to different features. The reason for the existence of multiple models is that for the same image, the features that need to be extracted may be multi-faceted, and these features may not be completely mutually exclusive. Therefore, it is difficult to pass the same model. extract these features. To this end, the feature extraction of the image of the commodity object can be realized from a variety of different angles through a plurality of different neural network models. For example, it may specifically include a model for distinguishing the area where the subject of the commodity object is located and the background area, a model for identifying whether the shooting scene is a street shot or a studio shot, and a model for identifying the display range of a character model. ,and many more. In short, for the same picture that needs feature extraction, it can be input into multiple different neural network models to extract features from different angles.

其中，具体实现时，所述神经网络模型可以是一些已有的模型，例如，现有技术中可能存在用于识别人脸的模型，等等。或者，在更多的情况下，可以是通过训练样本进行训练获得的，其中，所述训练样本中可以包括多个商品对象图片信息，以及根据图片点击率的影响因素对所述商品对象图片进行标注的标注信息。也即，通过具体的图片以及标注信息对模型进行训练，这样训练好的模型就可以用于识别该标注信息对应的特征。Wherein, during specific implementation, the neural network model may be some existing models, for example, a model for recognizing human faces may exist in the prior art, and so on. Or, in more cases, it may be obtained by training a training sample, wherein the training sample may include a plurality of commodity object picture information, and the commodity object picture may be processed according to the influencing factors of the picture click rate. Label information for the label. That is, the model is trained through specific pictures and annotation information, so that the trained model can be used to identify the features corresponding to the annotation information.

例如，具体的标注信息可以包括：训练样本图片中的商品对象主体所对应的区域以及背景所对应的区域。此时，在将具体商品对象的图片输入到训练好的模型中之后，所述神经网络模型输出层的输出结果可以包括：所述目标商品对象图片中多个像素属于商品对象主体或背景的类别信息。具体的，假设某商品对象的图片是一张100×100像素的图片，则该模型的输出可以是个100×100的矩阵，矩阵中每个元素的值可以为1或者0，其中，1可以表示对应的像素属于商品对象主体，0则表示对应的像素属于背景，等等。由于具体实现时，商品对象主体部分的主色调等信息会是影响用户点击率的因素，因此，在得到上述输出结果后，还可以根据所述输出结果确定所述元素类别中各像素的像素属性，通过比对各像素的像素属性确定目标像素属性，然后根据所述目标像素属性确定所述数据对象图片在至少一个维度上的特征值。例如，在一种具体的实现方式下，可以根据所述输出结果，确定所述目标商品对象图片中属于商品对象主体的多个像素的颜色属性信息，并确定所述商品对象主体部分的主色调信息，然后，可以将所述商品对象主体部分的主色调信息确定为所述特征向量中一维度上的特征值。For example, the specific annotation information may include: the area corresponding to the main body of the commodity object and the area corresponding to the background in the training sample picture. At this time, after the picture of the specific commodity object is input into the trained model, the output result of the output layer of the neural network model may include: a plurality of pixels in the picture of the target commodity object belong to the category of the main body or background of the commodity object information. Specifically, assuming that the picture of a commodity object is a picture of 100 × 100 pixels, the output of the model can be a 100 × 100 matrix, and the value of each element in the matrix can be 1 or 0, where 1 can represent The corresponding pixel belongs to the main body of the commodity object, 0 means that the corresponding pixel belongs to the background, and so on. Since information such as the main color of the main part of the commodity object will be a factor that affects the user's click-through rate during specific implementation, after the above output result is obtained, the pixel attribute of each pixel in the element category can also be determined according to the output result. , determining a target pixel attribute by comparing the pixel attributes of each pixel, and then determining a feature value of the data object picture in at least one dimension according to the target pixel attribute. For example, in a specific implementation manner, the color attribute information of a plurality of pixels belonging to the main body of the commodity object in the target commodity object picture may be determined according to the output result, and the main color of the main body of the commodity object may be determined Then, the main tone information of the main part of the commodity object may be determined as a feature value in one dimension in the feature vector.

或者，另一种方式下，还可以确定所述商品对象主体部分以及背景部分分别在所述目标商品对象图片中的占比，和/或连通区域的数量信息，这样，还可以将所述占比和/或所述连通区域的数量信息确定为所述特征向量中部分维度上的特征值，等等。Or, in another way, the proportion of the main part of the commodity object and the background part in the picture of the target commodity object, and/or the quantity information of the connected area, can also be determined. The ratio and/or the quantity information of the connected region is determined as the eigenvalues on some of the dimensions in the eigenvector, and so on.

另外，训练样本对应的标注信息也可以是：训练样本图片对应的拍摄场景类别信息；此时，所述神经网络模型输出层的输出结果包括：所述目标商品对象图片的拍摄场景所属的类别信息。具体的，所述拍摄场景类别信息也可以从多种不同角度具有不同的类别划分方式，例如，可以包括室内拍摄或室外拍摄，通过模特人物或道具对商品对象使用状态下进行的拍摄，或在无模特人物或道具情况下对商品对象细节进行的拍摄。其中，不同的类别划分方式下，可以使用不同的神经网络模型来进行特征提取。In addition, the labeling information corresponding to the training sample may also be: category information of the shooting scene corresponding to the training sample picture; in this case, the output result of the output layer of the neural network model includes: category information to which the shooting scene of the target commodity object picture belongs . Specifically, the category information of the shooting scene may also have different category division methods from various angles, for example, it may include indoor shooting or outdoor shooting, shooting of commodity objects through model characters or props in a state of use, or shooting in Detail shots of merchandise objects without mannequins or props. Among them, under different classification methods, different neural network models can be used for feature extraction.

再者，对于通过模特人物对商品对象使用状态进行展示的训练样本图片，具体的标注信息也可以包括：所述模特人物的特征信息。具体的，所述所述模特人物的特征信息同样可以从不同的角度具有多种不同的类别划分方式，例如，包括：所述目标商品对象图片中是否包括所述模特人物的全身/半身，或者，是否包括所述模特人物的人脸图像，或者，所述模特人物的姿态特征信息，或者，所述模特人物的表情特征信息，等等。类似的，不同划分方式下，可以使用不同的神经网络模型来进行特征提取。Furthermore, for the training sample pictures showing the usage state of the commodity object by the model character, the specific labeling information may also include: feature information of the model character. Specifically, the feature information of the model person may also have a variety of different classification methods from different angles, for example, including: whether the picture of the target product object includes the full body/half body of the model person, or , whether to include the face image of the model character, or the posture feature information of the model character, or the expression feature information of the model character, and so on. Similarly, under different division methods, different neural network models can be used for feature extraction.

S202：将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S202: Input the target commodity object image into the neural network model, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer represents the time when the output result is obtained. In the process, the feature information extracted by the middle layer of the neural network model from the picture of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

在获得了多个神经网络模型后，可以用这些模型对具体的目标商品对象的图片进行特征提取。其中，目标商品对象可以是需要向用户推荐的特定商品对象，例如，某商家新上架的一款新品，等等。或者，还可以是在需要向用户提供“猜你喜欢”等推荐信息的情况下，预先设定的商品池中的商品对象，等等。总之，对于每个可能会被推荐的具体的商品对象，都可以将商品对象的图片分别输入到多个神经网络模型中，进行特征的提取，以生成特征向量。其中，商品对象的图片具体可以是用作商品对象代表图的图片。所谓代表图就是指，在将商品对象投放到某页面中进行展示时，在该页面的资源位中展示出用于代表该商品对象的图片。After obtaining multiple neural network models, these models can be used to perform feature extraction on pictures of specific target commodity objects. The target commodity object may be a specific commodity object that needs to be recommended to the user, for example, a new product newly launched by a merchant, and the like. Alternatively, it may also be a commodity object in a preset commodity pool, and so on, when it is necessary to provide the user with recommendation information such as "guess you like it". In short, for each specific product object that may be recommended, the pictures of the product object can be input into multiple neural network models respectively, and features can be extracted to generate feature vectors. The picture of the commodity object may specifically be a picture used as a representative picture of the commodity object. The so-called representative image means that when a product object is placed on a page for display, a picture representing the product object is displayed in the resource position of the page.

其中，在本申请实施例中，在将具体的图片输入到神经网络模型后，具体获得的特征不仅可以包括神经网络模型输出层的输出结果，还可以包括中间层的激活值。也就是说，如图3所示，对于一个神经网络模型而言，通常会包括多层，每层中包括多个神经元。具体的，在卷积神经网络中，每个层可以称为一个卷积层，一个卷积层中，通常包含若干个特征平面(featureMap)，每个特征平面由一些矩形排列的的神经元组成。其中，在通过神经网络模型进行特征提取的过程中，每一个中间层的神经元也是分别在对图片进行特征提取，这些神经元提取出的特征可能是人类无法理解的一些抽象特征，但是，由于是与最终输出的特征判断相关的特征，因此，也会具有一定的参考价值。为此，在本申请实施例中，除了可以获得神经网络模型输出层的输出结果，还可以获得中间层的激活值，将这些激活值也可以作为描述图片特征的一部分。Wherein, in the embodiment of the present application, after a specific picture is input into the neural network model, the specific features obtained may include not only the output result of the output layer of the neural network model, but also the activation value of the intermediate layer. That is to say, as shown in Figure 3, a neural network model usually includes multiple layers, and each layer includes multiple neurons. Specifically, in a convolutional neural network, each layer can be called a convolutional layer, and a convolutional layer usually contains several feature planes (featureMap), and each feature plane is composed of some neurons arranged in a rectangle. . Among them, in the process of feature extraction through the neural network model, the neurons in each intermediate layer are also performing feature extraction on the picture, and the features extracted by these neurons may be some abstract features that humans cannot understand. However, due to It is a feature related to the feature judgment of the final output, so it will also have a certain reference value. For this reason, in the embodiment of the present application, in addition to the output results of the output layer of the neural network model, the activation values of the intermediate layers can also be obtained, and these activation values can also be used as part of describing the characteristics of the picture.

也就是说，在本申请实施例中，对于一个具体需要进行推荐等处理的商品对象的图片而言，可以输入到多个神经网络模型中，对于每个神经网络模型，都可以获得具体输出层的输出结果，以及至少一个中间层的激活值。That is to say, in the embodiment of the present application, for a picture of a commodity object that needs to be recommended and processed, it can be input into multiple neural network models, and for each neural network model, a specific output layer can be obtained. , and the activations of at least one intermediate layer.

其中，根据具体神经网络模型类别的不同，具体输出层的输出值以及中间层的激活值也存在不同。例如，在一种情况下，所述神经网络模型输出层的输出结果可以包括：所述目标商品对象图片中图像像素的元素类别特征信息，所述元素类别包括商品对象主体或背景，此时，所述中间层的激活值包括：在获得所述元素类别特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。或者，所述神经网络模型输出层的输出结果包括：所述目标商品对象图片的拍摄场景所属的类别特征信息；此时，所述中间层的激活值包括：在获得所述拍摄场景所属的类别特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。或者，所述神经网络模型输出层的输出结果也可以包括：对于通过模特人物对商品对象使用状态进行展示的训练样本图片，所述模特人物的特征信息；此时，所述中间层的激活值包括：在获得所述模特人物的特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。Among them, according to the different types of specific neural network models, the output value of the specific output layer and the activation value of the intermediate layer are also different. For example, in one case, the output result of the output layer of the neural network model may include: the element category feature information of the image pixels in the target commodity object picture, and the element category includes the commodity object main body or background, in this case, The activation value of the middle layer includes: in the process of obtaining the feature information of the element category, the feature information extracted by the middle layer of the neural network model from the picture of the target commodity object, or the feature information of the upper layer of neurons through the process. Feature information obtained by weighted summation and nonlinear transformation. Or, the output result of the output layer of the neural network model includes: the category feature information to which the shooting scene of the target commodity object picture belongs; at this time, the activation value of the intermediate layer includes: after obtaining the category to which the shooting scene belongs In the process of feature information, the middle layer of the neural network model extracts the feature information from the picture of the target commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer. Alternatively, the output result of the output layer of the neural network model may also include: for the training sample pictures showing the use status of the commodity object through the model character, the feature information of the model character; at this time, the activation value of the middle layer Including: in the process of obtaining the feature information of the model character, the middle layer of the neural network model extracts the feature information from the picture of the target commodity object, or weighted summation and nonlinearity of the neurons in the upper layer Transformed feature information.

S203：根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。S203: Determine feature values in multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer, and generate a feature vector of the commodity object image.

在获得具体神经网络模型的输出结果以及中间层的激活值后，可以分别确定为多个维度上的特征值，并进而生成商品对象图片的特征向量。其中，对于不同图片而言，具体在生成特征向量时，具体的维度信息可以是对应的。例如，第一维都可以表示图片的亮度信息，第二维表示图片中商品对象主体部分的主色调信息，第三维表示图片的拍摄场景类别，等等；而排序更靠后的一些维度中，则可能会一些抽象的特征，这些特征的维度标识可以用编号等进行代替，例如，第500维，第700维，等等。当然，这些以编号来表示的维度，具体可以对应同一神经网络的同一中间层中同一神经元的激活值。通过上述维度对齐，可以便于在不同图片的特征向量之间进行进行运算，例如，计算图片相似度等。After the output result of the specific neural network model and the activation value of the intermediate layer are obtained, they can be respectively determined as feature values in multiple dimensions, and then the feature vector of the image of the commodity object is generated. Wherein, for different pictures, when generating a feature vector, specific dimension information may be corresponding. For example, the first dimension can represent the brightness information of the picture, the second dimension can represent the main color information of the main part of the commodity object in the picture, the third dimension can represent the shooting scene category of the picture, etc.; Then there may be some abstract features, and the dimension identification of these features can be replaced by numbers, for example, the 500th dimension, the 700th dimension, and so on. Of course, these dimensions represented by numbers can specifically correspond to the activation values of the same neuron in the same intermediate layer of the same neural network. Through the above-mentioned dimensional alignment, it is convenient to perform operations between feature vectors of different pictures, for example, to calculate the similarity of pictures.

另外，具体实现时，还可以对所述输出层的输出结果以及所述中间层的激活值进行筛选，例如，在从多个模型中获得输出值以及中间层的激活值后，具体获得的特征数量可能会非常多，此时，还可以首先对这些特征进行筛选，从中过滤掉一些无用的特征，然后再根据筛选结果确定多个维度上的特征值，以此提高算法精度。In addition, during specific implementation, the output result of the output layer and the activation value of the intermediate layer can also be screened. For example, after obtaining the output value and the activation value of the intermediate layer from multiple models, the specific characteristics obtained The number may be very large. At this time, you can also filter these features first to filter out some useless features, and then determine the feature values in multiple dimensions according to the screening results, so as to improve the accuracy of the algorithm.

需要说明的是，在本申请实施例中，具体的商品对象图片可以包括商品对象关联的静态图片(照片等)，或者，还可以包括从商品对象关联的视频中提取出的图片，等等。对于前者，在具体向用户进行推荐时，可以将商品对象的图片推荐给用户关联的客户端，而对于后者，在通过视频中提取出的图片进行特征向量的生成以及与用户的匹配后，则可以将商品对象的视频推荐给用户关联的客户端。It should be noted that, in this embodiment of the present application, the specific picture of the commodity object may include a still picture (photo, etc.) associated with the commodity object, or may also include a picture extracted from a video associated with the commodity object, and so on. For the former, when specifically recommending to the user, the picture of the commodity object can be recommended to the client associated with the user, while for the latter, after the feature vector is generated and matched with the user through the picture extracted from the video, Then the video of the commodity object can be recommended to the client associated with the user.

总之，通过本申请实施例，能够通过神经网络模型来对商品对象图片进行特征提取，并且不仅可以将模型输出层的输出结果作为图片特征向量中具体维度上的特征值，还可以将模型中间层的激活值作为特征向量中更多维度上的特征值。这样，可以为商品对象图片生成高维度的特征向量，其中不仅可以包括人类可以理解、定义的维度上的特征值，还可以包括人类可能无法理解或者准确定义的维度上的特征值，从而使得这种特征向量能够更完整的表达图片的特征。进而在商品对象信息推荐等场景中，可以获得更准确的推荐结果，提高推荐信息的点击率，降低资源浪费。In a word, through the embodiments of the present application, it is possible to perform feature extraction on the image of the commodity object through the neural network model, and not only the output result of the model output layer can be used as the feature value of the specific dimension in the image feature vector, but also the intermediate layer of the model can be used. The activations of are used as eigenvalues in more dimensions in the eigenvector. In this way, a high-dimensional feature vector can be generated for the image of the commodity object, which can include not only the eigenvalues on the dimensions that humans can understand and define, but also the eigenvalues on the dimensions that humans may not understand or accurately define, so that the This kind of feature vector can express the features of the picture more completely. Furthermore, in scenarios such as product object information recommendation, more accurate recommendation results can be obtained, the click-through rate of the recommendation information can be improved, and the waste of resources can be reduced.

需要说明的是，本申请实施例中可能会涉及到对用户数据的使用，在实际应用中，可以在符合所在国的适用法律法规要求的情况下(例如，用户明确同意，对用户切实通知，等)，在适用法律法规允许的范围内在本文描述的方案中使用用户特定的个人数据。It should be noted that the embodiments of the present application may involve the use of user data. In practical applications, the user's data may be used in accordance with the applicable laws and regulations of the country where the user is located (for example, the user expressly agrees, and the user is effectively notified. etc.), use user-specific personal data in the scenarios described herein to the extent permitted by applicable laws and regulations.

实施例二Embodiment 2

上述实施例一提供了获得具体商品对象图片特征向量的方法，而在其他的应用场景中，也可以利用类似的方案来实现对用户特征向量的获取。其中，所谓的用户特征向量就是指用于描述用户特征的向量，这里的用户可以是指商品对象信息服务系统中的消费者用户、买家用户等。在向用户进行商品对象信息推荐，或者向用户推荐一些与其具有共同偏好的其他用户等场景中，通常需要将用户特征向量输入到具体的匹配算法中进行运算。其中，用户特征向量可以包括多个维度上的特征值，具体的维度可以包括用户的年龄、性别、职业等维度上的基本特征值，另外还可以根据用户历史行为(浏览、收藏、加购、购买等)相关的商品对象的信息，提取出用户感兴趣的商品对象所具有的共性等信息，这些维度上的信息也可以作为用户特征向量中具体维度上的特征值，用以对用户特征进行描述。The first embodiment above provides a method for obtaining a picture feature vector of a specific commodity object, and in other application scenarios, a similar solution can also be used to obtain the user feature vector. The so-called user feature vector refers to a vector used to describe user features, and the user here may refer to a consumer user, a buyer user, etc. in the commodity object information service system. In scenarios such as recommending commodity object information to a user, or recommending some other users with common preferences to the user, it is usually necessary to input the user feature vector into a specific matching algorithm for operation. The user feature vector may include feature values in multiple dimensions, and specific dimensions may include basic feature values in dimensions such as the user's age, gender, and occupation. Purchase, etc.) related commodity objects, and extract the commonalities and other information of commodity objects that the user is interested in. The information on these dimensions can also be used as the feature values of the specific dimensions in the user feature vector. describe.

其中，在从用户历史行为信息关联的商品对象中提取具体的商品对象特征信息时，除了可以包括商品对象的类目、品牌等基础信息外，同样还可以包括商品对象关联的图片的一些可视化特征，例如前文所述的拍摄场景类别，是否包含人物模特，等等。在获取这些可视化特征时，同样可以通过神经网络模型来获得，并且，同样可以获得神经网络模型输出层的输出结果以及中间层的激活值，从而生成用户的高维特征向量。其中可以包括人类可以理解并定义的维度上的特征值，也可以包括人类无法理解或者确切定义的维度上的特征值，以此实现对用户特征更全面更精细的描述。以使得推荐给用户的商品对象信息或者其他用户的信息更准确。Among them, when extracting specific commodity object feature information from commodity objects associated with user historical behavior information, in addition to basic information such as categories and brands of commodity objects, it can also include some visual features of pictures associated with commodity objects. , such as the above-mentioned shooting scene category, whether it contains character models, and so on. When obtaining these visual features, it can also be obtained through the neural network model, and the output results of the output layer of the neural network model and the activation value of the intermediate layer can also be obtained, so as to generate a high-dimensional feature vector of the user. It can include eigenvalues on dimensions that humans can understand and define, and can also include eigenvalues on dimensions that humans cannot understand or define exactly, so as to achieve a more comprehensive and detailed description of user characteristics. In order to make the commodity object information recommended to the user or the information of other users more accurate.

具体的，参见图4，该实施例二提供了一种确定用户特征信息的方法，该方法具体可以包括：Specifically, referring to FIG. 4 , the second embodiment provides a method for determining user feature information, and the method may specifically include:

S401：获取与用户的历史行为信息关联的多个商品对象的图片信息；S401: Acquire picture information of multiple commodity objects associated with the user's historical behavior information;

其中，商品对象的图片具体可以是指商品对象的代表图片，也就是说，在商品对象的详情页面中作为商品对象的头图的图片，或者，在投放到具体的商品对象列表页面时，在资源位中展示出的图片。由于用户如果从众多商品对象中对其中某个商品对象进行了点击，则该商品对象吸引该用户的因素除了其类目、品牌等，还可能在于该商品对象的代表图片引起了用户的兴趣。因此，这种商品对象图片的特征，也可以作为用户特征的一部分对用户特征进行描述。The picture of the commodity object may specifically refer to the representative picture of the commodity object, that is, the picture that is used as the header picture of the commodity object in the details page of the commodity object, or, when placed on the specific commodity object list page, in the The image displayed in the resource slot. Since a user clicks on a certain commodity object from among many commodity objects, besides its category, brand, etc., the factor that attracts the commodity object to the user may also be that the representative picture of the commodity object arouses the user's interest. Therefore, the feature of the commodity object image can also be used as a part of the user feature to describe the user feature.

S402：获取至少一个用于对商品对象图片进行特征提取的神经网络模型；S402: Acquire at least one neural network model for feature extraction on commodity object pictures;

S403：将多个商品对象的图片分别输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S403: Input pictures of a plurality of commodity objects into the neural network model respectively, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer indicates that after obtaining the In the process of outputting the result, the feature information extracted by the middle layer of the neural network model from the picture of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

S404：通过对所述多个商品对象的图片分别对应的输出层的输出结果以及所述中间层的激活值进行统计，确定多个维度上的特征值，以用于生成所述用户的特征向量。S404: Determine feature values in multiple dimensions by counting the output results of the output layers corresponding to the pictures of the multiple commodity objects and the activation values of the intermediate layers respectively, so as to generate the feature vector of the user .

在分别获得多个商品对象的图片对应的输出层的输出结果以及所述中间层的激活值后，可以进行统计，例如，对各种特征出现的数量或者频度等进行统计，确定出用户对对应维度上特征的感兴趣程度，然后将这些维度对应的感兴趣程度等信息作为特征值，生成具体的用户特征向量。After obtaining the output results of the output layer corresponding to the pictures of the multiple commodity objects and the activation value of the intermediate layer, statistics can be performed, for example, the number or frequency of occurrence of various features can be counted to determine the user's preference for Corresponding to the degree of interest of the feature on the dimension, and then using the information such as the degree of interest corresponding to these dimensions as the feature value to generate a specific user feature vector.

实施例三Embodiment 3

前述实施例一提供了获得具体商品对象图片特征向量的方法，而在确定出具体商品对象图片的特征向量之后，可以应用到具体的商品对象推荐等场景中。具体的，该实施例二就提供了其中一种具体的应用场景，该场景可以是通过客户端首页等提供的“猜你喜欢”等推荐信息的场景。也就是说，在用户访问客户端首页时，首页中可以提供“猜你喜欢”版块，在该版块中，可以提供多个资源位，每个资源位中可以关联具体推荐的商品对象信息。而这里推荐出的商品对象信息，通常是根据当前访问者用户自身的特征与商品对象库中的商品对象进行匹配后确定的。其中，商品对象的特征就可以同商品对象特征向量来描述，其中就可以包括具体商品对象图片对应的特征，而商品对象图片的特征则可以包括一些形象化的特征，另外还可以包括通过神经网络中间层的激活值获得的抽象特征，以此生成商品对象的高维特征向量。然后，再将商品对象的特征向量与用户向量输入到推荐算法中进行匹配运算，确定出具体适合推荐给当前用户的商品对象。其中，对于用户向量，可以采用现有技术中的向量表达方式，或者也可以采用前述实施例二中的高维特征向量来进行表达，等等。The foregoing Embodiment 1 provides a method for obtaining a feature vector of a picture of a specific commodity object, and after the feature vector of the picture of a specific commodity object is determined, it can be applied to scenarios such as recommendation of a specific commodity object. Specifically, the second embodiment provides one of the specific application scenarios, and the scenario may be a scenario in which recommendation information such as "guess you like it" is provided through the home page of the client. That is to say, when a user visits the homepage of the client, the homepage can provide a "guess you like it" section, in which multiple resource slots can be provided, and each resource slot can be associated with specific recommended product object information. The commodity object information recommended here is usually determined after matching the characteristics of the current visitor user with the commodity objects in the commodity object library. Among them, the characteristics of the commodity object can be described with the commodity object feature vector, which can include the characteristics corresponding to the specific commodity object pictures, and the characteristics of the commodity object pictures can include some visual features, and can also include the neural network. The abstract features obtained by the activation values of the intermediate layers are used to generate high-dimensional feature vectors of commodity objects. Then, the feature vector of the commodity object and the user vector are input into the recommendation algorithm for matching operation, and the commodity object that is specifically suitable to be recommended to the current user is determined. Wherein, for the user vector, the vector expression manner in the prior art may be used, or the high-dimensional feature vector in the foregoing second embodiment may also be used for expression, and so on.

具体的，参见图5，该实施例三提供了一种商品对象推荐方法，该方法具体可以包括：Specifically, referring to FIG. 5 , the third embodiment provides a method for recommending commodity objects, and the method may specifically include:

S501：确定商品对象的目标推荐用户，以及商品对象推荐信息的来源数据库；S501: Determine the target recommending user of the commodity object and the source database of the commodity object recommendation information;

其中，具体的目标推荐用户可以是待获得商品对象推荐信息的用户，具体可以有多种情况。例如，可以是对某目标页面发起访问请求的用户，如果该目标页面中存在“猜你喜欢”等版块，则对该目标页面发起的访问请求，也可以同时作为该用户发起的获取推荐信息的请求。具体的推荐信息来源数据库也可以根据具体的应用场景而定，例如，可以是系统的全量数据库，或者也可以在某些业务、商家的一些小范围的数据库等等。Wherein, the specific target recommending user may be the user who is to obtain the recommendation information of the commodity object, and there may be various specific situations. For example, it can be a user who initiates an access request to a target page. If there are sections such as "Guess you like it" in the target page, the access request initiated by the target page can also be used as the user's request to obtain recommendation information. ask. The specific recommendation information source database can also be determined according to the specific application scenario, for example, it can be the full database of the system, or it can also be some small-scale databases of certain businesses and merchants, and so on.

S502：获取所述用户的特征向量信息，以及所述来源数据库中多个商品对象的特征向量信息，所述商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S502: Acquire feature vector information of the user and feature vector information of multiple commodity objects in the source database, where the feature vectors of the commodity objects include feature values of pictures of commodity objects in multiple dimensions, and the The feature value includes: after inputting the picture of the commodity object into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer and the activation value of at least one intermediate layer; the intermediate layer The activation value is the feature information extracted from the picture of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the feature obtained by the weighted summation and nonlinear transformation of the neurons in the previous layer information;

S503：将所述用户的特征向量信息以及所述多个商品对象的特征向量信息输入到推荐算法中，以预测将所述商品对象推荐给所述用户后获得的点击率信息。S503: Input the feature vector information of the user and the feature vector information of the plurality of commodity objects into a recommendation algorithm to predict click rate information obtained after recommending the commodity objects to the user.

其中，具体的推荐算法不属于本申请实施例中关注的重点，因此，这里不再详述。The specific recommendation algorithm does not belong to the focus of the embodiments of the present application, and therefore will not be described in detail here.

在获得具体的点击率预测结果后，可以从所述多个商品对象中确定出点击率符合条件的至少一个商品对象，以便用于推荐给所述用户。After obtaining the specific CTR prediction result, at least one commodity object whose CTR meets the condition may be determined from the plurality of commodity objects, so as to be used for recommendation to the user.

实施例四Embodiment 4

该实施例四提供了另一种具体的推荐场景。具体的，可以在给定一种具体商品对象的情况下，从众多用户中预测出最有可能点击该商品对象的一个或者多个用户，并将该商品对象推荐给该用户。这种情况主要可以出现在商家的新品推荐等场景中，例如，某商家新上架一款商品对象，需要选择一些用户进行定向的推荐。此时，可以利用该商品对象的特征向量与多个用户的用户特征向量进行匹配，确定出将商品对象推荐给各用户时，被用户点击的概率信息。进而再确定具体向哪些用户推荐。具体的，参见图6，该实施例四提供了一种商品对象推荐方法，该方法具体可以包括：The fourth embodiment provides another specific recommendation scenario. Specifically, given a specific commodity object, one or more users who are most likely to click on the commodity object can be predicted from many users, and the commodity object is recommended to the user. This situation can mainly occur in scenarios such as new product recommendations of merchants. For example, a merchant has newly launched a product object and needs to select some users for targeted recommendation. At this time, the feature vector of the product object can be used to match the user feature vectors of multiple users to determine the probability information of the product object being clicked by the user when the product object is recommended to each user. Then determine which users to recommend to. Specifically, referring to FIG. 6 , the fourth embodiment provides a method for recommending commodity objects, and the method may specifically include:

S601：确定待推荐的商品对象的图片信息；S601: Determine the picture information of the commodity object to be recommended;

S602：获取所述商品对象的特征向量信息，所述商品对象的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S602: Acquire feature vector information of the product object, where the feature vector of the product object includes feature values of the picture in multiple dimensions, and the feature value includes: by inputting the picture of the product object into at least one After a neural network model is used for feature extraction of commodity object pictures, the output result of the output layer, and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the process of obtaining the output result of the neural network model The feature information extracted from the picture of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

S603：获取多个用户的特征向量信息；S603: Obtain feature vector information of multiple users;

S604：将所述商品对象的特征向量信息以及所述多个用户的特征向量信息输入到推荐算法中，以预测将所述待推荐的商品对象推荐给所述用户后获得的点击率信息。S604: Input the feature vector information of the commodity object and the feature vector information of the multiple users into a recommendation algorithm to predict click rate information obtained after recommending the commodity object to be recommended to the user.

具体实现时，还可以根据所预测出的点击率信息，将所述待推荐的商品对象信息推荐给点击率预测值符合条件的用户。During specific implementation, the information of the commodity object to be recommended may also be recommended to the users whose predicted value of the click rate meets the conditions according to the predicted click rate information.

实施例五Embodiment 5

除了可以通过将商品对象特征向量与用户的特征向量进行匹配，实现具体商品对象向用户的推荐之外，本申请实施例生成的商品对象特征向量还可以应用到商品对象与商品对象之间的相似度计算等场景中。例如，在一种场景下，某些商品对象列表页面中可以在多个资源位中展示出多个商品对象的信息，其中，在每个资源位中都可以通过“找相似”等操作选项，用户可以通过点击该操作选项获得与该资源位中的商品对象相似的更多其他商品对象的信息。此时，就会涉及到将该资源位中的商品对象的特征向量与其他更多商品对象的特征向量进行相似度计算，以确定商品对象之间的相似度。或者，当前用户在执行针对某商品对象的详情页面浏览，或者添加到待购买集合，或者收藏等操作时，系统可能也会提供“看了又看”功能，提供与用户操作的商品对象相似的更多商品对象信息供用户进行选择，等等。此时，也会涉及到商品对象与商品对象之间的相似度比较，具体同样可以通过商品对象的特征向量之间的运算来进行确定。而关于商品对象的特征向量，则同样可以采用本申请实施例中提供的方案，采用高维向量的方式来实现，其中可以包括通过神经网络模型的中间层激活值获得的抽象特征。具体的，该实施例五提供了一种商品对象的提供方法，参见图7，该方法具体可以包括：In addition to realizing the recommendation of specific commodity objects to users by matching the commodity object feature vector with the user's feature vector, the commodity object feature vector generated in the embodiment of the present application can also be applied to the similarity between commodity objects and commodity objects. degree calculation and other scenarios. For example, in one scenario, some commodity object list pages can display information about multiple commodity objects in multiple resource slots, and in each resource slot, you can use the "find similarities" and other operation options, The user can obtain the information of other commodity objects similar to the commodity object in the resource slot by clicking this operation option. At this time, the similarity calculation between the feature vector of the commodity object in the resource position and the feature vectors of other commodity objects is involved to determine the similarity between the commodity objects. Or, when the current user is browsing the details page of a certain commodity object, or adding it to the to-be-purchased collection, or bookmarking, etc., the system may also provide a "look and see" function, which provides similar information to the commodity object operated by the user. More product object information for users to choose from, and so on. At this time, the comparison of the similarity between the commodity object and the commodity object will also be involved, which can also be determined by the operation between the feature vectors of the commodity objects. As for the feature vector of the commodity object, the solution provided in the embodiment of the present application can also be used to realize the high-dimensional vector, which can include abstract features obtained through the activation value of the middle layer of the neural network model. Specifically, the fifth embodiment provides a method for providing a commodity object. Referring to FIG. 7 , the method may specifically include:

S701：在根据目标商品对象提供相似商品对象信息的过程中，获得所述目标商品对象的特征向量信息，以及来源数据库中多个商品对象的特征向量信息；其中，商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S701: In the process of providing similar commodity object information according to the target commodity object, obtain the feature vector information of the target commodity object and the feature vector information of multiple commodity objects in the source database; wherein, the feature vectors of the commodity objects include commodities The feature value of the picture of the object in multiple dimensions, the feature value includes: after inputting the picture of the commodity object into at least one neural network model for feature extraction of the picture of the commodity object, the output result of the output layer , and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the activation value of the upper layer of neurons. Feature information obtained by weighted summation and nonlinear transformation;

具体的，在展示包括有至少一个商品对象信息的目标页面的过程中，接收针对指定商品对象获得相似商品对象信息的请求，并将所述指定商品对象确定为所述目标商品对象。或者，接收对指定商品对象进行详情页面浏览，或者添加到待购买集合，或者收藏的操作请求，并将所述指定商品对象确定为所述目标商品对象。Specifically, in the process of displaying the target page including at least one commodity object information, a request for obtaining similar commodity object information for a specified commodity object is received, and the specified commodity object is determined as the target commodity object. Or, receiving an operation request for browsing the details page of the specified commodity object, or adding it to the to-be-purchased set, or bookmarking, and determining the specified commodity object as the target commodity object.

S702：通过将所述目标商品对象的特征向量信息与来源数据库中多个商品对象的特征向量信息进行相似性比对，提供与所述指定商品对象相似度符合条件的商品对象信息。S702: By comparing the feature vector information of the target product object with the feature vector information of multiple product objects in the source database, the product object information that meets the conditions of similarity with the specified product object is provided.

实施例六Embodiment 6

除了可以用于商品对象的推荐或者相似商品对象信息的提供等场景外，具体商品对象图片的高维特征向量信息还可以应用到对商品对象的代表图片进行优化的场景中。具体的，由于商品对象的代表图片通常可以展示到具体页面的资源位内，并且图片的特征可能会影响到用户的点击。例如，对于某用户而言，如果在资源位中展示的是商品对象的其中一个图片，可能并不会点击该商品对象，但是，如果换成另一张图片，则可能会点击该商品对象，等等。针对这种情况，在本申请实施例中，在将具体商品对象投放到某目标页面中时，可以为该商品对象关联多张可选的图片，其中，具体将哪张图片作为代表图片显示到具体的资源位中，则可以根据具体访问者用户的特征而定。也就是说，可以在接收到具体访问者用户对目标页面的访问请求后，通过用户的特征向量与商品对象各个图片的特征向量进行匹配运算，计算出展示哪张图片时，可以为商品对象获得更高的点击概率，进而便可以将这张图片作为该商品对象的代表图片展示到具体的资源位中。这样，可以实现在商品对象代表图层面上的“千人千面”，以此提升商品对象的点击率。其中，具体图片的特征向量便可以是本申请实施例中生成的包含有一些抽象特征的高维特征向量，而用户的特征向量可以是低维也可以是高维，以此可以实现更精准的匹配。In addition to being used in scenarios such as recommendation of commodity objects or provision of information on similar commodity objects, the high-dimensional feature vector information of specific commodity object pictures can also be applied to scenarios in which representative pictures of commodity objects are optimized. Specifically, because the representative picture of the commodity object can usually be displayed in the resource position of a specific page, and the characteristics of the picture may affect the user's click. For example, for a user, if one of the pictures of the commodity object is displayed in the resource slot, the commodity object may not be clicked, but if it is replaced by another picture, the commodity object may be clicked. and many more. In view of this situation, in the embodiment of the present application, when a specific commodity object is placed on a certain target page, a plurality of optional pictures can be associated with the commodity object, and which picture is specifically displayed as a representative picture to the specific commodity object. In the resource position, it can be determined according to the characteristics of specific visitors and users. That is to say, after receiving a specific visitor user's access request to the target page, the user's feature vector and the feature vector of each image of the product object can be matched to calculate which image to display, which can be obtained for the product object. The higher the probability of clicks, the picture can be displayed in a specific resource position as a representative picture of the commodity object. In this way, "thousands of people and thousands of faces" can be realized on the surface of the product object representative layer, so as to improve the click-through rate of the product object. The feature vector of a specific picture may be a high-dimensional feature vector including some abstract features generated in the embodiment of the present application, and the feature vector of the user may be low-dimensional or high-dimensional, so that a more accurate match.

具体的，该实施例六提供了一种提供商品对象信息的方法，参见图8，该方法具体可以包括：Specifically, the sixth embodiment provides a method for providing commodity object information. Referring to FIG. 8 , the method may specifically include:

S801：接收对目标页面的访问请求，所述目标页面关联有多个资源位，用于展示多个商品对象的信息，其中，所述商品对象关联有多张不同的图片；S801: Receive an access request to a target page, where the target page is associated with a plurality of resource bits for displaying information of a plurality of commodity objects, wherein the commodity objects are associated with a plurality of different pictures;

其中，目标页面具体可以有多种，例如，可以是客户端首页，或者一些活动会场的主页，频道主页，等等。There may be various specific target pages, for example, it may be the home page of the client, or the home page of some event venues, the channel home page, and so on.

S802：获取所述多张不同的图片分别对应的特征向量信息，以及访问者用户的特征向量信息；其中，所述图片对应的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；S802: Obtain the feature vector information corresponding to the multiple different pictures respectively, and the feature vector information of the visitor user; wherein, the feature vector corresponding to the picture includes the feature values of the picture in multiple dimensions, so the The feature value includes: the output result of the output layer and the activation value of at least one middle layer after inputting the picture into at least one neural network model for feature extraction of the commodity object picture; the activation value of the middle layer The value is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

S803：将所述访问者用户的特征向量信息以及所述商品对象的多张图片对应的特征向量信息分别输入到预测算法中，以预测将所述商品对象的多张图片展示给所述用户后分别获得的点击率信息；S803: Input the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into the prediction algorithm respectively, so as to predict that after the multiple pictures of the commodity object are displayed to the user CTR information obtained respectively;

S804：根据预测结果，将点击率符合条件的图片作为对应商品对象的代表图片展示在所述目标页面对应的资源位中。S804: According to the prediction result, a picture with a click-through rate that meets the conditions is displayed in a resource position corresponding to the target page as a representative picture of the corresponding commodity object.

其中，关于实施例二至实施例六中的未详述部分，可以参见前述实施例一中的记载，这里不再赘述。Wherein, for the parts not described in detail in the second embodiment to the sixth embodiment, reference may be made to the description in the foregoing embodiment one, and details are not repeated here.

实施例七Embodiment 7

如前文所述，通过神经网络模型或者其他的算法模型，可以对商品对象的图片进行分析，确定出商品对象的图片中关联的展示场景类别信息，例如，可以包括走秀场景，喝咖啡场景，室内棚拍场景，室外街拍场景，运动场景，等等。而对于用户而言，在获取具体的用户特征信息时，也可以获取到该用户感兴趣的展示场景信息。例如，可以通过对用户历史浏览过的商品对象的图片进行分析，将这些图片对应的展示场景信息进行分析，可以确定出用户感兴趣的展示场景信息，等等。因此，在具体向用户进行信息推荐时，还可以结合这种具体商品对象图片关联的展示场景类别信息进行推荐。具体的，在该实施例七中，提供了一种信息推荐方法，参见图9，该方法具体可以包括：As mentioned above, through the neural network model or other algorithm model, the picture of the commodity object can be analyzed to determine the category information of the display scene associated with the picture of the commodity object, for example, it can include catwalk scene, coffee drinking scene, indoor scene Studio shooting scenes, outdoor street shooting scenes, sports scenes, etc. For a user, when acquiring specific user feature information, it is also possible to acquire display scene information that the user is interested in. For example, by analyzing the pictures of commodity objects that the user has browsed in the past, and analyzing the display scene information corresponding to these pictures, the display scene information that the user is interested in can be determined, and so on. Therefore, when specifically recommending information to the user, the recommendation may also be performed in combination with the display scene category information associated with the specific commodity object picture. Specifically, in the seventh embodiment, an information recommendation method is provided. Referring to FIG. 9 , the method may specifically include:

S901：确定目标商品对象关联的图片；S901: Determine the picture associated with the target commodity object;

S902：通过对所述图片进行特征分析，确定所述图片中的商品对象对应的展示场景类别信息；S902: Determine the display scene category information corresponding to the commodity object in the picture by analyzing the characteristics of the picture;

其中，对图片进行特征分析时，可以利用预先训练完成的神经网络模型，或者普通的算法模型来进行，这里不再详述。Among them, when performing feature analysis on a picture, a pre-trained neural network model or a common algorithm model can be used to perform the feature analysis, which will not be described in detail here.

S903：确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；S903: Determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

关于用户特征的获取，前述实施例一中也有相应的记载，这里亦不再赘述。Regarding the acquisition of the user characteristics, there are also corresponding records in the foregoing Embodiment 1, which will not be repeated here.

S904：根据所述图片对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；S904: Determine a target user according to the display scene category information corresponding to the picture and the display scene category information that the user is interested in;

S905：将所述图片推荐到所述目标用户关联的客户端。S905: Recommend the picture to the client associated with the target user.

实施例八Embodiment 8

该实施例八与实施例七类似，具体推荐的对象不再是商品对象的图片，而是商品对象的视频。这种视频可以有多种来源，例如，可以包括商家用户进行拍摄并上传的视频，或者，还可以是在直播场景中截取的视频，等等。在这种情况下，确定出待推荐的视频后，可以从视频中提取出若干帧图像，通过对图像特征进行分析，可以确定出视频对应的展示场景类别信息。进而可以将该视频推荐给对该场景类别感兴趣的用户。具体的，参见图10，该实施例八具体提供了一种信息推荐方法，该方法具体可以包括：The eighth embodiment is similar to the seventh embodiment, and the specific recommended object is no longer a picture of the commodity object, but a video of the commodity object. This kind of video can come from various sources, for example, it can include a video shot and uploaded by a business user, or it can also be a video captured in a live broadcast scene, and so on. In this case, after determining the video to be recommended, several frames of images can be extracted from the video, and by analyzing the image features, the category information of the display scene corresponding to the video can be determined. The video can then be recommended to users who are interested in the scene category. Specifically, referring to FIG. 10 , the eighth embodiment specifically provides an information recommendation method, which may specifically include:

S1001：确定目标商品对象关联的视频；S1001: Determine the video associated with the target commodity object;

S1002：从所述视频中提取至少一帧图像，通过对所述图像进行特征分析，确定所述视频中的商品对象对应的展示场景类别信息；S1002: Extract at least one frame of image from the video, and determine the display scene category information corresponding to the commodity object in the video by performing feature analysis on the image;

S1003：确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；S1003: Determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

S1004：根据所述视频对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；S1004: Determine a target user according to the display scene category information corresponding to the video and the display scene category information that the user is interested in;

S1005：将所述视频推荐到所述目标用户关联的客户端。S1005: Recommend the video to the client associated with the target user.

与实施例一相对应，本申请实施例还提供了一种信息处理装置，参见图11，该装置具体可以包括：Corresponding to the first embodiment, the embodiment of the present application further provides an information processing apparatus, referring to FIG. 11 , the apparatus may specifically include:

模型获取单元1101，用于获取至少一个用于对商品对象图片进行特征提取的神经网络模型；A model obtaining unit 1101, configured to obtain at least one neural network model for feature extraction of commodity object pictures;

分析单元1102，用于将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Theanalysis unit 1102 is configured to input the target commodity object image into the neural network model, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer is represented in In the process of obtaining the output result, the feature information extracted by the middle layer of the neural network model from the picture of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

图片特征向量生成单元1103，用于根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。The picture featurevector generating unit 1103 is configured to determine feature values in multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer, and generate a feature vector of the commodity object picture.

其中，所述神经网络模型是通过训练样本进行训练获得的，所述训练样本中包括多个商品对象图片信息，以及根据图片点击率对所述商品对象图片进行标注的标注信息；所述标注信息包括：用于区分训练样本图片中的商品对象区域以及背景区域的标注信息，或者，训练样本图片对应的拍摄场景类别信息，或者，训练样本图片中包括的模特人物的特征信息。Wherein, the neural network model is obtained by training a training sample, and the training sample includes image information of a plurality of commodity objects, and annotation information for annotating the image of the commodity object according to the picture click rate; the annotation information It includes: the label information used to distinguish the commodity object area and the background area in the training sample picture, or the shooting scene category information corresponding to the training sample picture, or the feature information of the model person included in the training sample picture.

具体的，所述神经网络模型输出层的输出结果包括：所述目标商品对象图片中图像像素的元素类别特征信息，所述元素类别包括商品对象主体或背景；此时，所述中间层的激活值包括：在获得所述元素类别特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。Specifically, the output result of the output layer of the neural network model includes: the element category feature information of the image pixels in the target commodity object picture, and the element category includes the commodity object subject or background; at this time, the activation of the intermediate layer The value includes: in the process of obtaining the element category feature information, the feature information extracted by the middle layer of the neural network model from the target commodity object image, or the weighted summation and nonlinearity of the neurons in the upper layer. Transformed feature information.

具体实现时，该装置还可以包括：When specifically implemented, the device may further include:

像素属性确定子单元，用于根据所述输出结果确定所述元素类别中各像素的像素属性；a pixel attribute determination subunit, configured to determine the pixel attribute of each pixel in the element category according to the output result;

目标像素属性确定子单元，用于通过比对各像素的像素属性确定目标像素属性；The target pixel attribute determination subunit is used to determine the target pixel attribute by comparing the pixel attributes of each pixel;

此时，所述图片特征向量生成单元具体可以用于：根据所述目标像素属性确定所述数据对象图片在至少一个维度上的特征值。In this case, the picture feature vector generating unit may be specifically configured to: determine the feature value of the data object picture in at least one dimension according to the target pixel attribute.

具体的，所述像素属性确定子单元具体可以用于：根据所述输出结果，确定所述目标商品对象图片中商品对象主体对应的多个像素的颜色属性信息；Specifically, the pixel attribute determination subunit may be specifically configured to: determine, according to the output result, color attribute information of a plurality of pixels corresponding to the main body of the commodity object in the target commodity object image;

所述目标像素属性确定子单元具体可以用于：确定所述商品对象主体部分的主色调信息；The target pixel attribute determination subunit can be specifically used to: determine the main tone information of the main part of the commodity object;

所述图片特征向量生成单元具体可以用于：将所述商品对象主体部分的主色调信息确定为所述特征向量中一维度上的特征值。The picture feature vector generating unit may be specifically configured to: determine the main tone information of the main part of the commodity object as a feature value in one dimension in the feature vector.

或者，另一种方式下，所述目标像素属性确定子单元具体可以用于：确定所述商品对象主体部分以及背景部分分别在所述目标商品对象图片中的占比，和/或连通区域的数量信息；Or, in another way, the target pixel attribute determination subunit may be specifically used to: determine the proportion of the main part of the commodity object and the background part in the picture of the target commodity object, and/or the proportion of the connected area. quantity information;

所述图片特征向量生成单元具体可以用于：将所述占比和/或所述连通区域的数量信息确定为所述特征向量中部分维度上的特征值。The picture feature vector generating unit may be specifically configured to: determine the ratio and/or the quantity information of the connected regions as feature values on some dimensions in the feature vector.

另外，所述神经网络模型输出层的输出结果包括：所述目标商品对象图片的拍摄场景所属的类别特征信息；此时，所述中间层的激活值包括：在获得所述拍摄场景所属的类别特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。其中，所述拍摄场景类别信息：室内拍摄或室外拍摄，通过模特人物或道具对商品对象使用状态下进行的拍摄，或在无模特人物或道具情况下对商品对象细节进行的拍摄。In addition, the output result of the output layer of the neural network model includes: the category feature information to which the shooting scene of the target commodity object image belongs; at this time, the activation value of the intermediate layer includes: after obtaining the category to which the shooting scene belongs In the process of feature information, the middle layer of the neural network model extracts the feature information from the picture of the target commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer. Wherein, the shooting scene category information: indoor shooting or outdoor shooting, shooting of the commodity object using a model character or props, or shooting of the details of the commodity object without a model character or props.

或者，所述神经网络模型输出层的输出结果包括：对于通过模特人物对商品对象使用状态进行展示的训练样本图片，所述模特人物的特征信息；所述中间层的激活值包括：在获得所述模特人物的特征信息的过程中，所述神经网络模型的中间层从所述目标商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息。其中，所述模特人物的特征信息包括：所述目标商品对象图片中是否包括所述模特人物的全身/半身，或者，是否包括所述模特人物的人脸图像，或者，所述模特人物的姿态特征信息，或者，所述模特人物的表情特征信息。Or, the output result of the output layer of the neural network model includes: for the training sample picture showing the use state of the commodity object through the model character, the feature information of the model character; the activation value of the middle layer includes: after obtaining the obtained In the process of describing the feature information of the model person, the feature information extracted by the middle layer of the neural network model from the picture of the target commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer . Wherein, the characteristic information of the model person includes: whether the target product object image includes the full body/half body of the model person, or whether it includes a face image of the model person, or the posture of the model person Feature information, or, facial expression feature information of the model character.

具体实现时，该装置还可以包括：When specifically implemented, the device may further include:

特征筛选单元，用于对所述输出层的输出结果以及所述中间层的激活值进行筛选；a feature screening unit, used for screening the output result of the output layer and the activation value of the intermediate layer;

所述图片特征向量生成单元具体可以用于：根据筛选结果确定多个维度上的特征值。The picture feature vector generating unit may be specifically configured to: determine feature values in multiple dimensions according to the screening result.

与实施例二相对应，本申请实施例还提供了一种确定用户特征信息的装置，参见图12，该装置可以包括：Corresponding to Embodiment 2, this embodiment of the present application further provides an apparatus for determining user feature information. Referring to FIG. 12 , the apparatus may include:

图片信息确定单元1201，用于确定与用户的历史行为信息关联的多个商品对象的图片信息；The pictureinformation determining unit 1201 is configured to determine the picture information of a plurality of commodity objects associated with the user's historical behavior information;

模型获取单元1202，用于获取至少一个用于对商品对象图片进行特征提取的神经网络模型；Amodel obtaining unit 1202, configured to obtain at least one neural network model for feature extraction on the image of the commodity object;

分析单元1203，用于将多个商品对象的图片分别输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息；Theanalysis unit 1203 is configured to input the pictures of a plurality of commodity objects into the neural network model respectively, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation of the intermediate layer The value represents the feature information extracted from the image of the commodity object by the middle layer of the neural network model in the process of obtaining the output result;

用户特征向量生成单元1204，用于通过对所述多个商品对象的图片分别对应的输出层的输出结果以及所述中间层的激活值进行统计，确定多个维度上的特征值，以用于生成所述用户的特征向量。The user featurevector generating unit 1204 is configured to determine feature values in multiple dimensions by performing statistics on the output results of the output layers corresponding to the pictures of the multiple commodity objects and the activation values of the intermediate layers respectively, for use in A feature vector for the user is generated.

与实施例三相对应，本申请实施例还提供了一种商品对象信息推荐装置，参见图13，该装置可以包括：Corresponding to the three-phase embodiment, the embodiment of the present application further provides a product object information recommendation device. Referring to FIG. 13 , the device may include:

信息确定单元1301，用于确定商品对象的目标推荐用户，以及商品对象推荐信息的来源数据库；Aninformation determination unit 1301, configured to determine the target recommending user of the commodity object, and the source database of the commodity object recommendation information;

特征向量获取单元1302，用于获取所述用户的特征向量信息，以及所述来源数据库中多个商品对象的特征向量信息，所述商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；A featurevector obtaining unit 1302, configured to obtain the feature vector information of the user and the feature vector information of multiple commodity objects in the source database, where the feature vectors of the commodity objects include pictures of the commodity objects in multiple dimensions The feature value includes: after inputting the picture of the commodity object into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer, and the activation of at least one intermediate layer value; the activation value of the middle layer is the feature information extracted from the picture of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the weighted summation of the neurons in the upper layer and the Feature information obtained by nonlinear transformation;

预测单元1303，用于将所述用户的特征向量信息以及所述多个商品对象的特征向量信息输入到推荐算法中，以预测将所述商品对象推荐给所述用户后获得的点击率信息。Theprediction unit 1303 is configured to input the feature vector information of the user and the feature vector information of the plurality of commodity objects into a recommendation algorithm, so as to predict the click rate information obtained after recommending the commodity objects to the user.

具体实现时，该装置还可以包括：When specifically implemented, the device may further include:

推荐单元，用于根据所预测出的点击率信息，从所述多个商品对象中确定出点击率符合条件的至少一个商品对象，以推荐给所述用户。A recommending unit, configured to determine, from the plurality of commodity objects, at least one commodity object whose click rate meets a condition according to the predicted click-through rate information, so as to recommend it to the user.

与实施例四相对应，本申请实施例还提供了一种商品对象推荐装置，参见图14，该装置可以包括：Corresponding to the fourth embodiment, the embodiment of the present application also provides a product object recommendation device, see FIG. 14 , the device may include:

图片信息确定单元1401，用于确定待推荐的商品对象的图片信息；The picture information determining unit 1401 is used to determine the picture information of the commodity object to be recommended;

商品特征向量获取单元1402，用于获取所述商品对象的特征向量信息，所述商品对象的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；The product featurevector obtaining unit 1402 is configured to obtain feature vector information of the product object, where the feature vector of the product object includes feature values of the picture in multiple dimensions, and the feature values include: After the picture of the commodity object is input into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the neural network The feature information extracted by the model from the picture of the commodity object in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

用户特征向量获取单元1403，哟关于获取多个用户的特征向量信息；User featurevector acquisition unit 1403, about acquiring feature vector information of multiple users;

预测单元1404，用于将所述商品对象的特征向量信息以及所述多个用户的特征向量信息输入到推荐算法中，以预测将所述待推荐的商品对象推荐给所述用户后获得的点击率信息。Theprediction unit 1404 is configured to input the feature vector information of the commodity object and the feature vector information of the multiple users into the recommendation algorithm, so as to predict the click obtained after recommending the commodity object to be recommended to the user rate information.

具体实现时，还可以包括：In specific implementation, it can also include:

预测单元，用于根据预测的点击率信息，将所述待推荐的商品对象信息推荐给点击率预测值符合条件的用户。The predicting unit is configured to recommend the commodity object information to be recommended to the users whose predicted value of the click rate meets the conditions according to the predicted click rate information.

与实施例五相对应，本申请实施例还提供了一种商品对象的提供装置，参见图15，该装置可以包括：Corresponding to the fifth embodiment, the embodiment of the present application also provides a device for providing a commodity object. Referring to FIG. 15 , the device may include:

特征向量获取单元1501，用于在根据目标商品对象提供相似商品对象信息的过程中，获取所述目标商品对象的特征向量信息，以及来源数据库中多个商品对象的特征向量信息；其中，商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；The featurevector obtaining unit 1501 is configured to obtain the feature vector information of the target product object and the feature vector information of multiple product objects in the source database in the process of providing similar product object information according to the target product object; The feature vector of the product object includes the feature values of the picture of the product object in multiple dimensions, and the feature value includes: after inputting the picture of the product object into at least one neural network model for feature extraction of the product object picture , the output result of the output layer, and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or The feature information obtained by the weighted summation and nonlinear transformation of the neurons in the previous layer;

相似商品对象信息提供单元1502，用于通过将所述目标商品对象的特征向量信息与来源数据库中多个商品对象的特征向量信息进行相似性比对，提供与所述指定商品对象相似度符合条件的商品对象信息。Similar commodity objectinformation providing unit 1502, configured to provide a similarity matching condition with the specified commodity object by comparing the feature vector information of the target commodity object with the feature vector information of multiple commodity objects in the source database. product object information.

另外，该装置还可以包括：In addition, the device may also include:

第一请求接收单元，用于在展示包括有至少一个商品对象信息的目标页面的过程中，接收针对指定商品对象获得相似商品对象信息的请求，并将所述指定商品对象确定为所述目标商品对象。a first request receiving unit, configured to receive a request for obtaining similar commodity object information for a specified commodity object in the process of displaying a target page including at least one commodity object information, and determine the specified commodity object as the target commodity object.

或者，第二请求接收单元，用于接收对指定商品对象进行详情页面浏览、或者添加到待购买集合、或者收藏的操作请求，并将所述指定商品对象确定为所述目标商品对象。Or, the second request receiving unit is configured to receive an operation request for browsing the details page of the specified commodity object, or adding it to the to-be-purchased set, or favorite, and determining the specified commodity object as the target commodity object.

与实施例六相对应，本申请实施例还提供了一种提供商品对象信息的装置，参见图16，该装置可以包括：Corresponding to Embodiment 6, this embodiment of the present application further provides an apparatus for providing commodity object information. Referring to FIG. 16 , the apparatus may include:

访问请求接收单元1601，用于接收对目标页面的访问请求，所述目标页面关联有多个资源位，用于展示多个商品对象的信息，其中，所述商品对象关联有多张不同的图片；An access request receiving unit 1601, configured to receive an access request to a target page, the target page is associated with a plurality of resource bits for displaying information of a plurality of commodity objects, wherein the commodity objects are associated with a plurality of different pictures ;

特征向量获取单元1602，用于获取所述多张不同的图片分别对应的特征向量信息，以及访问者用户的特征向量信息；其中，所述图片对应的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；The featurevector obtaining unit 1602 is used to obtain the feature vector information corresponding to the multiple different pictures respectively, and the feature vector information of the visitor user; wherein, the feature vector corresponding to the picture includes the picture in multiple dimensions The eigenvalues above, the eigenvalues include: the output result of the output layer and the activation value of at least one intermediate layer after inputting the picture into at least one neural network model for feature extraction of the commodity object picture; The activation value of the middle layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or obtained by weighted summation and nonlinear transformation of the neurons in the upper layer. characteristic information;

预测单元1603，用于将所述访问者用户的特征向量信息以及所述商品对象的多张图片对应的特征向量信息分别输入到预测算法中，以预测将所述商品对象的多张图片展示给所述用户后分别获得的点击率信息；Theprediction unit 1603 is used to input the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into the prediction algorithm respectively, so as to predict that the multiple pictures of the commodity object will be displayed to the prediction algorithm. The click-through rate information obtained respectively by the user;

代表图片确定单元1604，用于根据预测结果，将点击率符合条件的图片作为对应商品对象的代表图片展示在所述目标页面对应的资源位中。The representativepicture determining unit 1604 is configured to, according to the prediction result, display the picture whose click rate meets the conditions as the representative picture corresponding to the commodity object in the resource position corresponding to the target page.

与实施例七相对应，本申请实施例还提供了一种信息推荐装置，参见图17，该装置可以包括：Corresponding to Embodiment 7, this embodiment of the present application further provides an information recommendation apparatus, referring to FIG. 17 , the apparatus may include:

图片确定单元1701，用于确定目标商品对象关联的图片；The picture determination unit 1701 is used to determine the picture associated with the target commodity object;

展示场景确定单元1702，用于通过对所述图片进行特征分析，确定所述图片中的商品对象对应的展示场景类别信息；The displayscene determination unit 1702 is configured to determine the display scene category information corresponding to the commodity object in the picture by analyzing the characteristics of the picture;

用户特征信息获取单元1703，用于确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；A user featureinformation acquisition unit 1703, configured to determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

目标用户确定单元1704，用于根据所述图片对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；A targetuser determination unit 1704, configured to determine the target user according to the display scene category information corresponding to the picture and the display scene category information that the user is interested in;

图片推荐单元1705，用于将所述图片推荐到所述目标用户关联的客户端。Thepicture recommending unit 1705 is configured to recommend the picture to the client associated with the target user.

与实施例八相对应，本申请实施例还提供了一种信息推荐装置，参见图18，该装置可以包括：Corresponding to the eighth embodiment, this embodiment of the present application further provides an information recommendation apparatus, referring to FIG. 18 , the apparatus may include:

视频确定单元1801，用于确定目标商品对象关联的视频；Avideo determination unit 1801, configured to determine the video associated with the target commodity object;

展示场景确定单元1802，用于从所述视频中提取至少一帧图像，通过对所述图像进行特征分析，确定所述视频中的商品对象对应的展示场景类别信息；The displayscene determination unit 1802 is configured to extract at least one frame of image from the video, and determine the display scene category information corresponding to the commodity object in the video by analyzing the characteristics of the image;

用户特征信息获取单元1803，用于确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；A user featureinformation acquisition unit 1803, configured to determine a user set, and obtain user feature information corresponding to multiple users in the user set, where the user feature information includes display scene category information that the user is interested in;

目标用户确定单元1804，用于根据所述视频对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；A targetuser determination unit 1804, configured to determine a target user according to the category information of the presentation scene corresponding to the video and the category information of the presentation scene that the user is interested in;

视频推荐单元1805，用于将所述视频推荐到所述目标用户关联的客户端。Avideo recommendation unit 1805, configured to recommend the video to the client associated with the target user.

另外，本申请实施例还提供了一种电子设备，包括：In addition, the embodiment of the present application also provides an electronic device, including:

一个或多个处理器；以及one or more processors; and

与所述一个或多个处理器关联的存储器，所述存储器用于存储程序指令,所述程序指令在被所述一个或多个处理器读取执行时,执行如下操作:A memory associated with the one or more processors, the memory is used to store program instructions, and the program instructions, when read and executed by the one or more processors, perform the following operations:

获取至少一个用于对商品对象图片进行特征提取的神经网络模型；Acquire at least one neural network model for feature extraction on commodity object images;

将目标商品对象图片输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Input the target commodity object image into the neural network model, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer represents the process of obtaining the output result The feature information extracted by the middle layer of the neural network model from the image of the commodity object, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

根据所述输出层的输出结果以及所述中间层的激活值，确定多个维度上的特征值，并生成所述商品对象图片的特征向量。According to the output result of the output layer and the activation value of the intermediate layer, the feature values in multiple dimensions are determined, and the feature vector of the commodity object image is generated.

或者，or,

确定与用户的历史行为信息关联的多个商品对象的图片信息；Determine the picture information of multiple commodity objects associated with the user's historical behavior information;

获取至少一个用于对商品对象图片进行特征提取的神经网络模型；Acquire at least one neural network model for feature extraction on commodity object images;

将多个商品对象的图片分别输入到所述神经网络模型中，获得所述神经网络模型输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值表征在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Input the pictures of a plurality of commodity objects into the neural network model respectively, and obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the intermediate layer indicates that the output result is obtained after the The feature information extracted by the middle layer of the neural network model from the picture of the commodity object in the process of , or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

通过对所述多个商品对象的图片分别对应的输出层的输出结果以及所述中间层的激活值进行统计，确定多个维度上的特征值，以用于生成所述用户的特征向量。By counting the output results of the output layer corresponding to the pictures of the multiple commodity objects and the activation values of the intermediate layer respectively, the feature values in multiple dimensions are determined to be used to generate the feature vector of the user.

或者，or,

确定商品对象的目标推荐用户，以及商品对象推荐信息的来源数据库；Determine the target recommended users of the product object, and the source database of the recommended information of the product object;

获取所述用户的特征向量信息，以及所述来源数据库中多个商品对象的特征向量信息，所述商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为在获得输出结果的过程中所述神经网络模型的中间层从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information of the user and the feature vector information of multiple commodity objects in the source database, where the feature vectors of the commodity objects include the feature values of the pictures of the commodity objects in multiple dimensions, and the feature values Including: after the picture of the commodity object is input into at least one neural network model for feature extraction of the commodity object picture, the output result of the output layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer The value is the feature information extracted from the picture of the commodity object by the middle layer of the neural network model in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

将所述用户的特征向量信息以及所述多个商品对象的特征向量信息输入到推荐算法中，以预测将所述商品对象推荐给所述用户后获得的点击率信息。The feature vector information of the user and the feature vector information of the plurality of commodity objects are input into the recommendation algorithm to predict the click rate information obtained after the commodity objects are recommended to the user.

或者，or,

确定待推荐的商品对象的图片信息；Determine the picture information of the commodity object to be recommended;

获得所述商品对象的特征向量信息，所述商品对象的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information of the product object, the feature vector of the product object includes feature values of the picture in multiple dimensions, and the feature value includes: by inputting the picture of the product object into at least one user After the neural network model for feature extraction is performed on the image of the commodity object, the output result of the output layer, and the activation value of at least one intermediate layer; the activation value of the intermediate layer is changed from the neural network model in the process of obtaining the output result. Feature information extracted from the picture of the commodity object, or feature information obtained by weighted summation and nonlinear transformation of the neurons in the upper layer;

获取多个用户的特征向量信息；Obtain feature vector information of multiple users;

将所述商品对象的特征向量信息以及所述多个用户的特征向量信息输入到推荐算法中，以预测将所述待推荐的商品对象推荐给所述用户后获得的点击率信息。The feature vector information of the commodity object and the feature vector information of the multiple users are input into the recommendation algorithm to predict the click rate information obtained after the commodity object to be recommended is recommended to the user.

或者，or,

在根据目标商品对象提供相似商品对象信息的过程中，获得所述目标商品对象的特征向量信息，以及来源数据库中多个商品对象的特征向量信息；其中，商品对象的特征向量中包括商品对象的图片在多个维度上的特征值，所述特征值包括：通过将所述商品对象的图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；In the process of providing similar commodity object information according to the target commodity object, the feature vector information of the target commodity object and the feature vector information of a plurality of commodity objects in the source database are obtained; The feature values of the picture in multiple dimensions, the feature values include: after the picture of the commodity object is input into at least one neural network model for feature extraction on the picture of the commodity object, the output result of the output layer, and The activation value of at least one middle layer; the activation value of the middle layer is the feature information extracted from the picture of the commodity object by the neural network model in the process of obtaining the output result, or the weighting of the neurons in the upper layer Summation and feature information obtained by nonlinear transformation;

通过将所述目标商品对象的特征向量信息与来源数据库中多个商品对象的特征向量信息进行相似性比对，提供与所述指定商品对象相似度符合条件的商品对象信息。By comparing the feature vector information of the target product object with the feature vector information of multiple product objects in the source database, the product object information that meets the conditions of similarity with the specified product object is provided.

或者，or,

接收对目标页面的访问请求，所述目标页面关联有多个资源位，用于展示多个商品对象的信息，其中，所述商品对象关联有多张不同的图片；receiving an access request to a target page, the target page is associated with a plurality of resource bits for displaying information of a plurality of commodity objects, wherein the commodity objects are associated with a plurality of different pictures;

获得所述多张不同的图片分别对应的特征向量信息，以及访问者用户的特征向量信息；其中，所述图片对应的特征向量中包括所述图片在多个维度上的特征值，所述特征值包括：通过将所述图片输入到至少一个用于对商品对象图片进行特征提取的神经网络模型后，输出层的输出结果，以及至少一个中间层的激活值；所述中间层的激活值为所述神经网络模型在获得输出结果的过程中从所述商品对象的图片中抽取的特征信息，或对上一层神经元经过加权求和以及非线性变换得到的特征信息；Obtain the feature vector information corresponding to the multiple different pictures respectively, and the feature vector information of the visitor user; wherein, the feature vector corresponding to the picture includes the feature values of the picture in multiple dimensions, and the feature The value includes: the output result of the output layer and the activation value of at least one middle layer after inputting the picture into at least one neural network model for feature extraction of the commodity object picture; the activation value of the middle layer is The feature information extracted by the neural network model from the picture of the commodity object in the process of obtaining the output result, or the feature information obtained by the weighted summation and nonlinear transformation of the neurons in the upper layer;

将所述访问者用户的特征向量信息以及所述商品对象的多张图片对应的特征向量信息分别输入到预测算法中，以预测将所述商品对象的多张图片展示给所述用户后分别获得的点击率信息；The feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object are respectively input into the prediction algorithm, so as to predict that the multiple pictures of the commodity object are displayed to the user and obtained respectively. click-through rate information;

根据预测结果，将点击率符合条件的图片作为对应商品对象的代表图片展示在所述目标页面对应的资源位中。According to the prediction result, a picture whose click rate meets the conditions is displayed in the resource position corresponding to the target page as a representative picture of the corresponding commodity object.

或者，or,

确定目标商品对象关联的图片；Determine the picture associated with the target product object;

通过对所述图片进行特征分析，确定所述图片中的商品对象对应的展示场景类别信息；Determine the display scene category information corresponding to the commodity object in the picture by analyzing the characteristics of the picture;

确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；determining a set of users, and acquiring user feature information corresponding to multiple users in the user set, where the user feature information includes category information of display scenarios that the user is interested in;

根据所述图片对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；Determine the target user according to the display scene category information corresponding to the picture and the display scene category information that the user is interested in;

将所述图片推荐到所述目标用户关联的客户端。The picture is recommended to the client associated with the target user.

或者，or,

确定目标商品对象关联的视频；Determine the video associated with the target product object;

从所述视频中提取至少一帧图像，通过对所述图像进行特征分析，确定所述视频中的商品对象对应的展示场景类别信息；Extract at least one frame of image from the video, and determine the display scene category information corresponding to the commodity object in the video by analyzing the characteristics of the image;

确定一用户集合，并获取所述用户集合中的多个用户分别对应的用户特征信息，所述用户特征信息包括用户感兴趣的展示场景类别信息；determining a set of users, and acquiring user feature information corresponding to multiple users in the user set, where the user feature information includes category information of display scenarios that the user is interested in;

根据所述视频对应的展示场景类别信息，以及所述用户感兴趣的展示场景类别信息，确定目标用户；Determine the target user according to the display scene category information corresponding to the video and the display scene category information that the user is interested in;

将所述视频推荐到所述目标用户关联的客户端。The video is recommended to the client associated with the target user.

其中，图19示例性的展示出了电子设备的架构，具体可以包括处理器1910，视频显示适配器1911，磁盘驱动器1912，输入/输出接口1913，网络接口1914，以及存储器1920。上述处理器1910、视频显示适配器1911、磁盘驱动器1912、输入/输出接口1913、网络接口1914，与存储器1920之间可以通过通信总线1930进行通信连接。19 exemplarily shows the architecture of the electronic device, which may specifically include a processor 1910 , a video display adapter 1911 , a disk drive 1912 , an input/output interface 1913 , a network interface 1914 , and a memory 1920 . The processor 1910 , the video display adapter 1911 , the disk drive 1912 , the input/output interface 1913 , and the network interface 1914 , and the memory 1920 can be communicatively connected through thecommunication bus 1930 .

其中，处理器1910可以采用通用的CPU(Central Processing Unit，中央处理器)、微处理器、应用专用集成电路(Application Specific Integrated Circuit，ASIC)、或者一个或多个集成电路等方式实现，用于执行相关程序，以实现本申请所提供的技术方案。The processor 1910 may be implemented by a general-purpose CPU (Central Processing Unit, central processing unit), a microprocessor, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used for Execute relevant programs to realize the technical solutions provided by this application.

存储器1920可以采用ROM(Read Only Memory，只读存储器)、RAM(Random AccessMemory，随机存取存储器)、静态存储设备，动态存储设备等形式实现。存储器1920可以存储用于控制电子设备1900运行的操作系统1921，用于控制电子设备1900的低级别操作的基本输入输出系统(BIOS)。另外，还可以存储网页浏览器1923，数据存储管理系统1924，以及信息处理系统1925等等。上述信息处理系统1925就可以是本申请实施例中具体实现前述各步骤操作的应用程序。总之，在通过软件或者固件来实现本申请所提供的技术方案时，相关的程序代码保存在存储器1920中，并由处理器1910来调用执行。The memory 1920 may be implemented in the form of a ROM (Read Only Memory, read only memory), a RAM (Random Access Memory, random access memory), a static storage device, a dynamic storage device, and the like. The memory 1920 may store an operating system 1921 for controlling the operation of the electronic device 1900 , a basic input output system (BIOS) for controlling low-level operations of the electronic device 1900 . In addition, a web browser 1923, a datastorage management system 1924, an information processing system 1925, and the like may also be stored. The above-mentioned information processing system 1925 may be an application program that specifically implements the operations of the foregoing steps in this embodiment of the present application. In a word, when the technical solutions provided in the present application are implemented through software or firmware, the relevant program codes are stored in the memory 1920 and invoked by the processor 1910 for execution.

输入/输出接口1913用于连接输入/输出模块，以实现信息输入及输出。输入输出/模块可以作为组件配置在设备中(图中未示出)，也可以外接于设备以提供相应功能。其中输入设备可以包括键盘、鼠标、触摸屏、麦克风、各类传感器等，输出设备可以包括显示器、扬声器、振动器、指示灯等。The input/output interface 1913 is used to connect input/output modules to realize information input and output. The input/output/module can be configured in the device as a component (not shown in the figure), or can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, and the like.

网络接口1914用于连接通信模块(图中未示出)，以实现本设备与其他设备的通信交互。其中通信模块可以通过有线方式(例如USB、网线等)实现通信，也可以通过无线方式(例如移动网络、WIFI、蓝牙等)实现通信。The network interface 1914 is used to connect a communication module (not shown in the figure), so as to realize the communication interaction between the device and other devices. The communication module may implement communication through wired means (eg, USB, network cable, etc.), or may implement communication through wireless means (eg, mobile network, WIFI, Bluetooth, etc.).

总线1930包括一通路，在设备的各个组件(例如处理器1910、视频显示适配器1911、磁盘驱动器1912、输入/输出接口1913、网络接口1914，与存储器1920)之间传输信息。Bus 1930 includes a path that transfers information between the various components of the device (eg, processor 1910, video display adapter 1911, disk drive 1912, input/output interface 1913, network interface 1914, and memory 1920).

需要说明的是，尽管上述设备仅示出了处理器1910、视频显示适配器1911、磁盘驱动器1912、输入/输出接口1913、网络接口1914，存储器1920，总线1930等，但是在具体实施过程中，该设备还可以包括实现正常运行所必需的其他组件。此外，本领域的技术人员可以理解的是，上述设备中也可以仅包含实现本申请方案所必需的组件，而不必包含图中所示的全部组件。It should be noted that although the above device only shows the processor 1910, the video display adapter 1911, the disk drive 1912, the input/output interface 1913, the network interface 1914, the memory 1920, thebus 1930, etc., in the specific implementation process, the A device may also include other components necessary for proper operation. In addition, those skilled in the art can understand that, the above-mentioned device may also include only the necessary components to realize the solution of the present application, instead of all the components shown in the figures.

通过以上的实施方式的描述可知，本领域的技术人员可以清楚地了解到本申请可借助软件加必需的通用硬件平台的方式来实现。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在存储介质中，如ROM/RAM、磁碟、光盘等，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本申请各个实施例或者实施例的某些部分所述的方法。From the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products can be stored in storage media, such as ROM/RAM, magnetic disks , CD-ROM, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present application.

本说明书中的各个实施例均采用递进的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于系统或系统实施例而言，由于其基本相似于方法实施例，所以描述得比较简单，相关之处参见方法实施例的部分说明即可。以上所描述的系统及系统实施例仅仅是示意性的，其中所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性劳动的情况下，即可以理解并实施。Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and reference may be made to the partial description of the method embodiment for related parts. The systems and system embodiments described above are only illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, It can be located in one place, or it can be distributed over multiple network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

以上对本申请所提供的信息处理方法、装置及电子设备，进行了详细介绍，本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处。综上所述，本说明书内容不应理解为对本申请的限制。The information processing method, device and electronic device provided by the present application have been introduced in detail above. The principles and implementations of the present application are described with specific examples in this paper. The descriptions of the above embodiments are only used to help understand the present application. At the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation and application scope. In conclusion, the content of this specification should not be construed as a limitation on the present application.

Claims (38)

1. An information processing method characterized by comprising:

acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

inputting a target commodity object picture into the neural network model to obtain an output result of an output layer of the neural network model and an activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and determining characteristic values on multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer, and generating a characteristic vector of the commodity object picture.

2. The method of claim 1,

the neural network model is obtained by training a training sample, wherein the training sample comprises a plurality of commodity object picture information and marking information for marking the commodity object picture according to the picture click rate;

the labeling information includes: and the marking information is used for distinguishing the commodity object area and the background area in the training sample picture, or shooting scene type information corresponding to the training sample picture, or characteristic information of a model figure included in the training sample picture.

3. The method of claim 1,

the output result of the neural network model output layer comprises: element category feature information of image pixels in the target commodity object picture, wherein the element category comprises a commodity object main body or a background;

the activation values of the intermediate layer include: in the process of obtaining the element category feature information, the intermediate layer of the neural network model extracts feature information from the target commodity object picture, or obtains feature information by performing weighted summation and nonlinear transformation on neurons in the upper layer.

4. The method of claim 3, further comprising:

determining the pixel attribute of each pixel in the element category according to the output result;

determining a target pixel attribute by comparing pixel attributes of the pixels; the determining the feature values in the multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer includes:

and determining the characteristic value of the data object picture on at least one dimension according to the target pixel attribute.

5. The method of claim 4,

the determining the pixel attribute of each pixel in the element category according to the output result includes:

determining color attribute information of a plurality of pixels corresponding to the commodity object main body in the target commodity object picture according to the output result;

the determining the target pixel attribute by comparing the pixel attributes of the pixels includes:

determining dominant hue information of the merchandise object body portion;

the determining a feature value of the data object picture in at least one dimension according to the target pixel attribute comprises:

and determining the main tone information of the main body part of the commodity object as a characteristic value in one dimension in the characteristic vector.

6. The method of claim 3,

the determining the target pixel attribute by comparing the pixel attributes of the pixels includes:

determining the proportion of the main body part of the commodity object and the background part of the commodity object in the target commodity object picture and/or the quantity information of the connected areas;

the determining a feature value of the data object picture in at least one dimension according to the target pixel attribute comprises:

and determining the occupation ratio and/or the quantity information of the connected regions as characteristic values on partial dimensions in the characteristic vector.

7. The method of claim 1,

the output result of the neural network model output layer comprises: the category characteristic information of the shooting scene of the target commodity object picture;

the activation values of the intermediate layer include: in the process of obtaining the category characteristic information of the shooting scene, the intermediate layer of the neural network model extracts the characteristic information from the target commodity object picture or obtains the characteristic information obtained by weighting summation and nonlinear transformation on the neuron in the previous layer.

8. The method of claim 7,

the shooting scene category information: indoor shooting or outdoor shooting, shooting of commodity object using states through model characters or props, or shooting of commodity object details without the model characters or props.

9. The method of claim 1,

the output result of the neural network model output layer comprises: for training sample pictures for displaying the using state of the commodity object through a model figure, the characteristic information of the model figure;

the activation values of the intermediate layer include: in the process of obtaining the feature information of the model person, the intermediate layer of the neural network model extracts the feature information from the target commodity object picture or obtains the feature information by performing weighted summation and nonlinear transformation on neurons in the previous layer.

10. The method of claim 9,

the characteristic information of the model character comprises: whether the whole body/half body of the model character is included in the target commodity object picture, whether a face image of the model character is included in the target commodity object picture, or whether the posture characteristic information of the model character is included in the target commodity object picture, or whether the facial image of the model character is included in the target commodity object picture, or whether the posture characteristic information of the model character is included in the target commodity object picture, or whether the expression characteristic information of the model.

11. The method of any one of claims 1 to 10, further comprising:

screening the output result of the output layer and the activation value of the intermediate layer;

the determining feature values in a plurality of dimensions includes:

and determining characteristic values on multiple dimensions according to the screening result.

12. A method for determining user characteristic information, comprising:

determining picture information of a plurality of commodity objects associated with historical behavior information of a user;

acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

respectively inputting pictures of a plurality of commodity objects into the neural network model to obtain an output result of an output layer of the neural network model and an activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and determining characteristic values on multiple dimensions by counting the output results of the output layers respectively corresponding to the pictures of the commodity objects and the activation values of the middle layer so as to generate the characteristic vector of the user.

13. A method for recommending a commodity object, comprising:

determining a target recommendation user of the commodity object and a source database of the commodity object recommendation information;

acquiring feature vector information of the user and feature vector information of a plurality of commodity objects in the source database, wherein the feature vectors of the commodity objects comprise feature values of pictures of the commodity objects in a plurality of dimensions, and the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the characteristic information extracted from the picture of the commodity object by the intermediate layer of the neural network model in the process of obtaining the output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on the neuron on the upper layer;

and inputting the feature vector information of the user and the feature vector information of the plurality of commodity objects into a recommendation algorithm to predict click rate information obtained after recommending the commodity objects to the user.

14. The method of claim 13, further comprising:

and determining at least one commodity object with a click rate meeting the condition from the plurality of commodity objects according to the predicted click rate information so as to recommend the commodity object to the user.

15. A method for recommending a commodity object, comprising:

determining picture information of a commodity object to be recommended;

acquiring feature vector information of the commodity object, wherein the feature vector of the commodity object comprises feature values of the picture in multiple dimensions, and the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

acquiring feature vector information of a plurality of users;

and inputting the characteristic vector information of the commodity object and the characteristic vector information of the users into a recommendation algorithm to predict click rate information obtained after the commodity object to be recommended is recommended to the users.

16. The method of claim 15, further comprising:

and recommending the commodity object information to be recommended to users with click rate predicted values meeting conditions according to the predicted click rate information.

17. A method for providing a commodity object, comprising:

in the process of providing similar commodity object information according to a target commodity object, acquiring feature vector information of the target commodity object and feature vector information of a plurality of commodity objects in a source database; the feature vector of the commodity object comprises feature values of pictures of the commodity object in multiple dimensions, wherein the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

and comparing the similarity of the characteristic vector information of the target commodity object with the characteristic vector information of a plurality of commodity objects in a source database to provide commodity object information with the similarity meeting the condition of the specified commodity object.

18. The method of claim 17, further comprising, prior to the method:

in the process of displaying a target page including at least one commodity object information, receiving a request for obtaining similar commodity object information for a specified commodity object, and determining the specified commodity object as the target commodity object.

19. The method of claim 17, further comprising, prior to the method:

receiving an operation request for browsing a detail page or adding to a to-be-purchased set or collecting a specified commodity object, and determining the specified commodity object as the target commodity object.

20. A method of providing merchandise object information, comprising:

receiving an access request to a target page, wherein the target page is associated with a plurality of resource positions and is used for displaying information of a plurality of commodity objects, and the commodity objects are associated with a plurality of different pictures;

acquiring feature vector information corresponding to the multiple different pictures and feature vector information of the visitor user; wherein, the feature vector corresponding to the picture comprises feature values of the picture in multiple dimensions, and the feature values comprise: after the pictures are input into at least one neural network model for carrying out feature extraction on the commodity object pictures, outputting the output result of the layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

respectively inputting the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into a prediction algorithm so as to predict click rate information respectively obtained after the multiple pictures of the commodity object are displayed to the user;

and according to the prediction result, displaying the picture with the click rate meeting the condition as a representative picture of the corresponding commodity object in the resource position corresponding to the target page.

21. An information recommendation method, comprising:

determining a picture associated with the target commodity object;

determining display scene category information corresponding to the commodity object in the picture by performing feature analysis on the picture;

determining a user set, and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, wherein the user characteristic information comprises display scene category information which is interesting to the users;

determining a target user according to the display scene category information corresponding to the picture and the display scene category information interested by the user;

and recommending the picture to a client associated with the target user.

22. An information recommendation method, comprising:

determining a video associated with the target commodity object;

extracting at least one frame of image from the video, and determining display scene category information corresponding to the commodity object in the video by performing feature analysis on the image;

determining a user set, and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, wherein the user characteristic information comprises display scene category information which is interesting to the users;

determining a target user according to the display scene category information corresponding to the video and the display scene category information which is interested by the user;

and recommending the video to the client associated with the target user.

23. An information processing apparatus characterized by comprising:

the model acquisition unit is used for acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

the analysis unit is used for inputting the target commodity object picture into the neural network model to obtain an output result of an output layer of the neural network model and an activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and the picture characteristic vector generating unit is used for determining characteristic values on a plurality of dimensions according to the output result of the output layer and the activation value of the intermediate layer and generating the characteristic vector of the commodity object picture.

24. An apparatus for determining user characteristic information, comprising:

the picture information determining unit is used for determining picture information of a plurality of commodity objects related to historical behavior information of the user;

the model acquisition unit is used for acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

the analysis unit is used for respectively inputting the pictures of the plurality of commodity objects into the neural network model to obtain the output result of the output layer of the neural network model and the activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and the user characteristic vector generating unit is used for determining characteristic values on a plurality of dimensions by counting the output results of the output layers respectively corresponding to the pictures of the commodity objects and the activation values of the middle layer so as to generate the characteristic vector of the user.

25. A commodity object information recommendation device characterized by comprising:

the information determining unit is used for determining a target recommending user of the commodity object and a source database of the commodity object recommending information;

a feature vector obtaining unit, configured to obtain feature vector information of the user and feature vector information of a plurality of commodity objects in the source database, where the feature vector of a commodity object includes feature values of pictures of the commodity object in multiple dimensions, and the feature values include: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the characteristic information extracted from the picture of the commodity object by the intermediate layer of the neural network model in the process of obtaining the output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on the neuron on the upper layer;

and the prediction unit is used for inputting the characteristic vector information of the user and the characteristic vector information of the plurality of commodity objects into a recommendation algorithm so as to predict click rate information obtained after the commodity objects are recommended to the user.

26. A commodity object recommending apparatus, comprising:

the picture information determining unit is used for determining the picture information of the commodity object to be recommended;

a commodity feature vector obtaining unit, configured to obtain feature vector information of the commodity object, where the feature vector of the commodity object includes feature values of the picture in multiple dimensions, and the feature values include: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

a user feature vector acquisition unit having feature vector information on acquiring a plurality of users;

and the prediction unit is used for inputting the characteristic vector information of the commodity object and the characteristic vector information of the users into a recommendation algorithm so as to predict click rate information obtained after the commodity object to be recommended is recommended to the users.

27. An apparatus for providing a commodity object, comprising:

the characteristic vector acquisition unit is used for acquiring characteristic vector information of a target commodity object and characteristic vector information of a plurality of commodity objects in a source database in the process of providing similar commodity object information according to the target commodity object; the feature vector of the commodity object comprises feature values of pictures of the commodity object in multiple dimensions, wherein the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

and the similar commodity object information providing unit is used for comparing the similarity of the characteristic vector information of the target commodity object with the characteristic vector information of a plurality of commodity objects in a source database, and providing the commodity object information with the similarity meeting the condition of the specified commodity object.

28. An apparatus for providing commodity object information, comprising:

the system comprises an access request receiving unit, a processing unit and a display unit, wherein the access request receiving unit is used for receiving an access request to a target page, the target page is associated with a plurality of resource positions and is used for displaying information of a plurality of commodity objects, and the commodity objects are associated with a plurality of different pictures;

the characteristic vector acquisition unit is used for acquiring characteristic vector information corresponding to the different pictures and characteristic vector information of the visitor user; wherein, the feature vector corresponding to the picture comprises feature values of the picture in multiple dimensions, and the feature values comprise: after the pictures are input into at least one neural network model for carrying out feature extraction on the commodity object pictures, outputting the output result of the layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

the prediction unit is used for respectively inputting the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into a prediction algorithm so as to predict click rate information respectively obtained after the multiple pictures of the commodity object are displayed to the user;

and the representative picture determining unit is used for displaying the picture with the click rate meeting the conditions as the representative picture of the corresponding commodity object in the resource position corresponding to the target page according to the prediction result.

29. An information recommendation apparatus, comprising:

the picture determining unit is used for determining a picture associated with the target commodity object;

the display scene determining unit is used for determining display scene category information corresponding to the commodity object in the picture by performing characteristic analysis on the picture;

the system comprises a user characteristic information acquisition unit, a display unit and a display unit, wherein the user characteristic information acquisition unit is used for determining a user set and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, and the user characteristic information comprises display scene category information which is interesting to the users;

the target user determining unit is used for determining a target user according to the display scene category information corresponding to the picture and the display scene category information which is interested by the user;

and the picture recommending unit is used for recommending the picture to the client associated with the target user.

30. An information recommendation apparatus, comprising:

the video determining unit is used for determining a video associated with the target commodity object;

the display scene determining unit is used for extracting at least one frame of image from the video and determining display scene category information corresponding to commodity objects in the video by performing feature analysis on the image;

the system comprises a user characteristic information acquisition unit, a display unit and a display unit, wherein the user characteristic information acquisition unit is used for determining a user set and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, and the user characteristic information comprises display scene category information which is interesting to the users;

the target user determining unit is used for determining a target user according to the display scene category information corresponding to the video and the display scene category information which is interested by the user;

and the video recommending unit is used for recommending the video to the client associated with the target user.

31. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

inputting a target commodity object picture into the neural network model to obtain an output result of an output layer of the neural network model and an activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and determining characteristic values on multiple dimensions according to the output result of the output layer and the activation value of the intermediate layer, and generating a characteristic vector of the commodity object picture.

32. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

determining picture information of a plurality of commodity objects associated with historical behavior information of a user;

acquiring at least one neural network model for carrying out feature extraction on the commodity object picture;

respectively inputting pictures of a plurality of commodity objects into the neural network model to obtain an output result of an output layer of the neural network model and an activation value of at least one intermediate layer; the activation value of the middle layer represents feature information extracted from the commodity object picture by the middle layer of the neural network model in the process of obtaining an output result, or feature information obtained by performing weighted summation and nonlinear transformation on neurons in the upper layer;

and determining characteristic values on multiple dimensions by counting the output results of the output layers respectively corresponding to the pictures of the commodity objects and the activation values of the middle layer so as to generate the characteristic vector of the user.

33. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

determining a target recommendation user of the commodity object and a source database of the commodity object recommendation information;

acquiring feature vector information of the user and feature vector information of a plurality of commodity objects in the source database, wherein the feature vectors of the commodity objects comprise feature values of pictures of the commodity objects in a plurality of dimensions, and the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the intermediate layer is the characteristic information extracted from the picture of the commodity object by the intermediate layer of the neural network model in the process of obtaining the output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on the neuron on the upper layer;

and inputting the feature vector information of the user and the feature vector information of the plurality of commodity objects into a recommendation algorithm to predict click rate information obtained after recommending the commodity objects to the user.

34. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

determining picture information of a commodity object to be recommended;

obtaining feature vector information of the commodity object, wherein the feature vector of the commodity object comprises feature values of the picture in multiple dimensions, and the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

acquiring feature vector information of a plurality of users;

and inputting the characteristic vector information of the commodity object and the characteristic vector information of the users into a recommendation algorithm to predict click rate information obtained after the commodity object to be recommended is recommended to the users.

35. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

in the process of providing similar commodity object information according to a target commodity object, obtaining feature vector information of the target commodity object and feature vector information of a plurality of commodity objects in a source database; the feature vector of the commodity object comprises feature values of pictures of the commodity object in multiple dimensions, wherein the feature values comprise: after the pictures of the commodity objects are input into at least one neural network model for carrying out feature extraction on the pictures of the commodity objects, outputting the output result of a layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

and comparing the similarity of the characteristic vector information of the target commodity object with the characteristic vector information of a plurality of commodity objects in a source database to provide commodity object information with the similarity meeting the condition of the specified commodity object.

36. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

receiving an access request to a target page, wherein the target page is associated with a plurality of resource positions and is used for displaying information of a plurality of commodity objects, and the commodity objects are associated with a plurality of different pictures;

obtaining feature vector information corresponding to the multiple different pictures and feature vector information of the visitor user; wherein, the feature vector corresponding to the picture comprises feature values of the picture in multiple dimensions, and the feature values comprise: after the pictures are input into at least one neural network model for carrying out feature extraction on the commodity object pictures, outputting the output result of the layer and the activation value of at least one intermediate layer; the activation value of the middle layer is the characteristic information extracted from the picture of the commodity object by the neural network model in the process of obtaining an output result, or the characteristic information obtained by carrying out weighted summation and nonlinear transformation on neurons in the upper layer;

respectively inputting the feature vector information of the visitor user and the feature vector information corresponding to the multiple pictures of the commodity object into a prediction algorithm so as to predict click rate information respectively obtained after the multiple pictures of the commodity object are displayed to the user;

and according to the prediction result, displaying the picture with the click rate meeting the condition as a representative picture of the corresponding commodity object in the resource position corresponding to the target page.

37. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

determining a picture associated with the target commodity object;

determining display scene category information corresponding to the commodity object in the picture by performing feature analysis on the picture;

determining a user set, and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, wherein the user characteristic information comprises display scene category information which is interesting to the users;

determining a target user according to the display scene category information corresponding to the picture and the display scene category information interested by the user;

and recommending the picture to a client associated with the target user.

38. An electronic device, comprising:

one or more processors; and

a memory associated with the one or more processors for storing program instructions that, when read and executed by the one or more processors, perform operations comprising:

determining a video associated with the target commodity object;

extracting at least one frame of image from the video, and determining display scene category information corresponding to the commodity object in the video by performing feature analysis on the image;

determining a user set, and acquiring user characteristic information corresponding to a plurality of users in the user set respectively, wherein the user characteristic information comprises display scene category information which is interesting to the users;

determining a target user according to the display scene category information corresponding to the video and the display scene category information which is interested by the user;

and recommending the video to the client associated with the target user.

Images