A weakly supervised fine-grained image recognition method based on visual self-attention mechanism

Abstract

The invention discloses a weakly supervised fine-grained image recognition method based on a visual self-attention mechanism, including student-model, teacher-model and classification-model modules; the student-model and teacher-model are learned by Pairwise Approach-based Teacher ‑Student loop feedback mechanism is combined to form a self-attention region recommendation network, thereby enhancing the connection between discriminative region positioning and fine-grained feature learning, and can still detect fine-grained features more accurately in the absence of target bounding boxes and part label points. The discriminative area in the granular image significantly improves the recognition accuracy; at the same time, the three modules of student-model, teacher-model and classification-model share the convolutional layer, effectively compressing the model storage space and reducing the calculation cost, making the method meet real-time It is suitable for large-scale real scenarios; in addition, the dynamic weight distribution mechanism is used in multi-task joint learning to reduce the amount of artificially set hyperparameters and enhance the robustness of the model. Finally, the overall model is trained in an end-to-end single-stage manner and learning to reduce the difficulty of network optimization.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a weakly supervised fine-grained image recognition method based on a visual self-attention mechanism. Background technique [0002] Fine-grained image recognition is a challenging research topic in the field of computer vision, the purpose is to distinguish different subcategories under the same category. Compared with the cross-species coarse-grained image recognition task, the appearance similarity of different types of targets in the fine-grained image dataset is relatively high, while the same type of target has significant visual differences due to factors such as illumination, posture, and viewing angle, so the general depth It is difficult for learning image recognition techniques to accurately discriminate fine-grained object categories. Previous studies have shown that the difficulty of fine-grained image recognition tasks lies in discriminative region localizat...

AI Technical Summary

Problems solved by technology

This training method is more complicated, and it is easy to fall into a local optimal solution.

[0005] In summary, for weakly supervised fine-grained image recognition tasks that only use image-level label information, existing methods are difficult to detect image discriminative regions and learn fine-grained features simply, accurately and stably, so an adaptive A fine-grained image recognition method based on visual self-attention mechanism with high robustness

Images