A fine‐grained image classification method based on information interaction

• DOI: 10.1049/ipr2.13295
• Publication Date: 2024-12-04T06:39:21Z
• Authors: Shuo Zhu; Xukang Zhang; Yu Wang; Zongyang Wang; Jiahao Sun
• Keywords: Contextual image classification

Abstract To enhance the accuracy of fine‐grained image classification and address challenges such as excessive interference factors within dataset, inadequate extraction local key features, insufficient channel semantic association, a dual‐branch information interaction model that integrates convolutional neural networks (CNN) with Vision Transformers is proposed. This leverages Transformer branch to extract global which are subsequently combined CNN further augment model's capability for extraction. In order ability reduce loss feature information, enhancement module added branch. Since directly convolves convolution kernel will result in inability learn underlying features image, shallow proposed, branches interact dual through down‐sampling Down up‐sampling UP module. The improved method on CUB‐200‐2011, Stanford Cars FGVC‐Aircraft datasets 95.2%, 97.1% 96.9%, respectively. experimental results show has good generalization different datasets.