Fine-Grained Image Classification Based on a Collaborative Dual-Path Attention Contrastive Network

View at publisher Summarise with VIRAT
Authors: Jiayi Liu, Yue Su, Junbo Wang, Mingjun Qiu, Mingcheng Liu
Cited by: 0
References: 6

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Fine-grained image classification is a fundamental challenge in computer vision, aiming to distinguish subtle differences among subcategories within the same parent class. Existing methods under weak supervision struggle to effectively capture multiple complementary discriminative regions. This paper proposes a Collaborative Dual-path Attention Contrastive Network (CDACN), which adopts ResNet-50 as the backbone. A dual-path attention mechanism with differential initialization is introduced to localize distinct salient regions, while feature-decoupled contrastive learning is employed to enhance feature representation. An adaptive gated fusion module is further designed to integrate heterogeneous features. Experiments on the CUB-200-2011 dataset demonstrate that CDACN achieves a Top-1 classification accuracy of 84.3%, significantly outperforming the ResNet-50 baseline (78.5%). The interpretability of the model is reflected in attention region visualization, explicit constraints imposed by feature decoupling, and the analyzability of dynamic fusion weights. This method provides an efficient and interpretable solution for fine-grained image recognition.

Fields of Study & Subjects

Advanced Neural Network Applications Domain Adaptation and Few-Shot Learning Advanced Image and Video Retrieval Techniques

Publication details

Journal / source
International Conference on Artificial Intelligence and Social Network Systems (AISNS)
Pages
84–87
Publication date
9 January 2026
Language
EN