A Fine-grained Image Recognition Method Based on Feature Comparison and Channel Attention Mechanism

Abstract

The invention belongs to the field of deep learning, computer vision and fine-grained image classification, and specifically relates to a fine-grained image recognition method, system and device based on a feature comparison-based channel attention mechanism, aiming at solving the problem of less marked data in fine-grained images. In the case of the conventional model, the recognition accuracy rate is low, and the structure of the fine-grained classification model is complex. The invention includes: extracting the feature map of the sample and obtaining the basic feature vector after nonlinear mapping and average pooling; calculating and updating the category average feature vector, and comparing it with the sample basic feature vector; encoding the comparison result; according to the sample basic feature vector Learn the basic attention weight of the feature channel; fuse the encoding result and the basic attention weight and map to obtain the final attention weight to guide the model training; apply the trained model to fine-grained image recognition. The invention improves the classification accuracy of conventional classifiers in fine-grained tasks in a simple and effective manner.

Description

technical field [0001] The invention belongs to the fields of deep learning, computer vision and fine-grained image classification, and specifically relates to a fine-grained image recognition method, system and device based on a feature comparison-based channel attention mechanism. Background technique [0002] Image classification is a classic and important task in computer vision. With the great success of deep learning in computer vision in recent years, the task of image classification has made great progress. On this basis, more and more specific application scenarios need to be specially studied. In some scenarios, image recognition between similar objects has very important application value. For example, in the study of birds, identifying the species of bird is often the first step in the research. If birds with similar appearances but different species can be accurately and automatically identified, it can reduce the effort of scholars in bird identification, wh...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the above problems in the prior art, that is, in the case of less fine-grained image annotation data, the recognition accuracy of the conventional model is low, and the structure of the fine-grained classification model is complex, the present invention provides a feature-based comparison A fine-grained image recognition method of channel attention mechanism, the fine-grained image recognition method includes:

Images