Cross-Hierarchical Bidirectional Consistency Learning for Fine-Grained Visual Classification

2025-04-18

AI Summary

Fine-grained visual classification (FGVC) suffers from small inter-class variance and large intra-class variance, while existing methods often neglect the structured semantic relationships encoded in hierarchical tree-structured labels. To address this, we propose Cross-Hierarchical Bidirectional Consistency learning (CHBC), the first framework to jointly model both parent-to-child and child-to-parent hierarchical constraints. CHBC employs attention mask decomposition enhancement, multi-level feature disentanglement, and a bidirectional consistency loss to enable joint top-down and bottom-up predictive calibration. Evaluated on three standard benchmarks—CUB-200-2011, Stanford Cars, and FGVC-Aircraft—CHBC achieves state-of-the-art performance. Ablation studies demonstrate that its components synergistically improve classification accuracy by 2.3% and reduce misclassification rate by 18.7%, significantly alleviating fine-grained confusion.

Technology Category

• Computer Vision: Visual Reasoning & Symbolic Representations
• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Fine-Grained Visual Classification (FGVC) aims to categorize closely related subclasses, a task complicated by minimal inter-class differences and significant intra-class variance. Existing methods often rely on additional annotations for image classification, overlooking the valuable information embedded in Tree Hierarchies that depict hierarchical label relationships. To leverage this knowledge to improve classification accuracy and consistency, we propose a novel Cross-Hierarchical Bidirectional Consistency Learning (CHBC) framework. The CHBC framework extracts discriminative features across various hierarchies using a specially designed module to decompose and enhance attention masks and features. We employ bidirectional consistency loss to regulate the classification outcomes across different hierarchies, ensuring label prediction consistency and reducing misclassification. Experiments on three widely used FGVC datasets validate the effectiveness of the CHBC framework. Ablation studies further investigate the application strategies of feature enhancement and consistency constraints, underscoring the significant contributions of the proposed modules.

Problem

• Classify closely related subclasses with minimal inter-class differences
• Utilize Tree Hierarchies to improve classification accuracy and consistency
• Reduce misclassification via bidirectional consistency loss across hierarchies

Innovation

• Cross-Hierarchical Bidirectional Consistency Learning framework
• Decomposes and enhances attention masks and features
• Employs bidirectional consistency loss for label consistency