Image understanding method and system based on self-learning attributes

Abstract

The invention relates to an image understanding method and system based on self-learning attributes, which belong to the technical field of computer vision and image understanding. The image understanding system comprises an input module, an attribute label generation module, a convolutional neural network module and an output module. The convolutional neural network module comprises a backbone network, an image understanding task model and a self-learning attribute model; in the training stage of the method, a training sample A is decoded and preprocessed to obtain an image matrix and a tasklabel; the decoded training sample A is expanded to obtain a sample B and a sample C, an attribute label is generated based on the relationship among the sample A, the sample B and the sample C, and then an optimal model parameter is acquired; in a test stage, the test image is decoded and preprocessed to obtain an image matrix, and the input test image is predicted to obtain an image understanding result. According to the method, extra manual annotation is not needed, features with better representation capability can be obtained, and the result of image understanding is further improved.

Description

technical field [0001] The invention relates to an image understanding method and system based on self-learning attributes, belonging to the technical field of image understanding. Background technique [0002] Image understanding refers to the use of computer systems to analyze image data input into the system and extract descriptive information that can be understood by humans. Typical image understanding tasks include image recognition, object detection, scene understanding, etc. With the development of deep learning, image understanding methods based on convolutional neural networks have gradually become the mainstream. These methods mine effective image features from training samples, and then map the features to specific image understanding task spaces. [0003] However, there are still many problems in image understanding methods based on convolutional neural networks. First of all, these methods are based on a large number of training samples, and there is a large ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to propose an image understanding method and system based on self-study attributes in view of the technical defects of existing image understanding methods such as large training sample requirements, difficulty in manual labeling, and interference features in the background. Introduce self-learning attribute constraints into the image understanding model to extract better image features, thereby improving the accuracy of image understanding tasks

Images