Fine-grained image classification algorithm based on discriminant learning

Abstract

The invention belongs to the technical field of computer vision, and provides a fine-grained image classification algorithm based on discriminant learning. A new end-to-end autoregressive positioningand discriminative priori network model is provided. The new end-to-end autoregressive positioning and discriminative priori network model discriminates the size of the patch more accurately through learning and exploring, and can classify images in real time. Specifically, a multi-task discriminant learning network is designed, and the multi-task discriminant learning network comprises an autoregressive positioning sub-network and a discriminant prior sub-network, and the discriminant prior sub-network has a guide loss function and a consistency loss function to learn an autoregressive coefficient and a discriminant prior map at the same time. The autoregressive coefficient can reduce noise information in the discriminative patch, and thousands of candidate patches are filtered into the number of bits by learning the discriminative probability value by the discriminative priori map. A large number of experiments show that the proposed SDN model reaches the latest level in accuracy andefficiency.

Description

technical field [0001] The invention belongs to the technical field of computer vision, starts from improving the accuracy and efficiency of fine-grained image classification, and proposes a fine-grained image classification algorithm based on discriminant learning. Background technique [0002] Different from general image classification, weakly supervised fine-grained image classification (WFGIC) only uses image-level labels to identify objects at more detailed categories and granularity. Due to its numerous potential applications in image understanding and computer vision systems, WFGIC has received extensive attention from academia and industry. WFGIC is an open problem in the field of computer vision, not only because images belonging to the same subcategory can vary widely in size, pose, color, and background, while images belonging to different subcategories can be very similar in these aspects, and only using It is very difficult to extract discriminative features f...

AI Technical Summary

Problems solved by technology

A large number of about 2000 candidate patches generated by SS need to be predicted whether to contain discriminative features through a deep CNN classification network, which requires high computational cost

Images