Fine-grained image recognition method and device, storage medium and program product

By performing multi-scale feature extraction and channel attention feature enhancement on fine-grained images, we generate reinforced images and input recognition models, and solve the problem of low accuracy when extracting image detail features based on Transformer model, achieving higher recognition accuracy.

CN119649056APending Publication Date: 2025-03-18长江三峡(海南)绿色发展投资有限公司 +1
0 Cites 1 Cited by

Patent Information

Application Number
CN202411648432.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-18
Publication Date
2025-03-18

AI Technical Summary

Technical Problem

When extracting image detail features, the fine-grained image recognition method based on the Transformer model cannot fully capture the key local features required to distinguish different types of images, resulting in low recognition accuracy.

Method used

Multi-scale feature extraction of fine-grained images to be recognized can be obtained by obtaining multiple feature maps; then these feature maps are enhanced to channel attention feature enhancement to generate attention masks; key features in the image are enhanced based on the attention mask to form an enhanced image; finally, the enhanced image is input to the image recognition model for identification.

Benefits of technology

This method can fully capture the key local features required to distinguish different types of images, significantly improving the accuracy of fine-grained image recognition.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a fine-grained image recognition method and device, a storage medium and a program product, and the method comprises the steps: carrying out the multi-scale feature extraction of a to-be-recognized fine-grained image, and obtaining a plurality of feature maps of the to-be-recognized fine-grained image; channel attention feature enhancement is carried out on the plurality of feature maps of the fine-grained image to be recognized, an attention mask is obtained, and the attention mask can represent the key degree of each pixel in the fine-grained image to be recognized; on the basis of the attention mask, key features in the fine-grained image are enhanced, and an enhanced image with the key features enhanced is obtained; and inputting the enhanced image into the image recognition model for image recognition to obtain the recognition result of the fine-grained image, thereby improving the accuracy of the recognition degree of the fine-grained image.
Need to check novelty before this filing date? Find Prior Art