Adaptive Feature Alignment and Enhancement for Precise Fine-Grained Visual Recognition

Fine-grained visual classification (FGVC) faces challenges in distinguishing subtle differences among visually similar categories. This paper introduces a novel adaptive part-aware feature alignment and enhancement network (PAAE) designed to effectively extract discriminative features. Our approach comprises a Progressive Part Mining Module for capturing discriminative features across different layers, an Adaptive Scale Displacement Alignment Module for addressing feature space misalignment, and a Dual-Path Feature Enhancement Module for highlighting foreground features. Experimental results on benchmark datasets demonstrate competitive performance, with an accuracy of 92.4% on the CUB-200-2011 dataset and 95.1% on the Stanford Dogs dataset, showcasing the efficacy of our proposed method. The source code will be made publicly available at https://github.com/XubaozZ/PAAE.