Fine-Grained Visual Classification (FGVC) poses significant challenges due to subtle inter-class differences and large intra-class variations. Existing approaches often rely on feature-selection mechanisms or region-proposal strategies to localize discriminative regions for semantic analysis. However, these methods frequently fail to comprehensively capture discriminative cues while introducing substantial category-agnostic redundancy. To address these limitations, we propose H3Former, a novel token-to-region framework that leverages high-order semantic relations to aggregate local fine-grained representations with structured region-level modeling.
Specifically, we introduce the Semantic-Aware Aggregation Module (SAAM), which exploits multi-scale contextual cues to dynamically construct a weighted hypergraph among tokens. By applying hypergraph convolution, SAAM captures high-order semantic dependencies and progressively aggregates token features into compact region-level representations.
Furthermore, we present the Hyperbolic Hierarchical Contrastive Loss (HHCL), enforcing hierarchical semantic constraints in a non-Euclidean embedding space. HHCL enhances inter-class separability and intra-class consistency while preserving intrinsic hierarchical relationships among fine-grained categories.
Comprehensive experiments conducted on four standard FGVC benchmarks validate the superiority of our H3Former framework.
Blogger's Review: H3Former significantly enhances fine-grained visual classification performance by incorporating high-order semantic relations and hypergraph convolution. This innovative aggregation method effectively reduces redundancy and improves the model's discriminative ability, showcasing the potential of hierarchical contrast in non-Euclidean spaces, making it a noteworthy contribution to the field.