Image feature acquisition method and device, and electronic equipment
A technology of image features and acquisition methods, applied in the computer field, can solve problems such as weak image expression ability
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0027] A method for acquiring image features disclosed in this embodiment, such as figure 1 As shown, the method includes: Step 100 to Step 140.
[0028] Step 100, training a classification model by using training images of preset categories.
[0029] Optionally, in the embodiment of the present application, a classification model is trained based on a deep convolutional neural network model. A classification model is a model for identifying product categories based on product images. When training a classification model, a large number of product images of various categories are usually required as training images. The product images mentioned in the examples in this application may be images of dishes on the ordering platform, clothing images on the clothing sales platform, or scenic spots on the travel consultation platform. During specific implementation, the format of the training image of the model based on the deep convolutional neural network is usually (label, image...
Embodiment 2
[0042] Based on the first embodiment, this embodiment discloses a specific implementation manner of an image feature acquisition method.
[0043] Preferably, the training the classification model by using the training images of the preset category includes: based on the idea of maximizing the variance between classes, training the classification model based on the deep convolutional neural network by using the training images of the preset category.
[0044] During specific implementation, 5,000 types of training images can be taken, wherein each type of sample includes 10,000 product images. After manually setting the category label for each image, the image with the category label will be set to generate training data. For example, the format is: (label, image ) training data as the input of the deep convolutional neural network. At the same time, a multi-task model based on a deep convolutional neural network is constructed based on the idea of maximizing variance betwe...
Embodiment 3
[0073] An image feature acquisition device disclosed in this embodiment, such as image 3 As shown, the device includes:
[0074] Classification model training module 300, for training the classification model by training images of preset categories;
[0075] The dissimilar image pair determination module 310 is used to test the classification result of the classification model trained by the classification model training module 300 by verifying the image, and determine the dissimilar image pair confused by the classification model;
[0076] A similar image pair determination module 320, configured to determine a similar image pair based on the training images of the preset category;
[0077] A classification model optimization module 330, configured to optimize the classification model based on the similar image pair and the dissimilar image pair;
[0078] The image feature acquisition module 340 uses the optimized classification model to acquire image features.
[0079] o...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More