A Relational Model for Fine-Grained Visual Classification

Abstract

Fine-grained visual classification is challenging due to subtle inter-class differences and strong visual similarity among categories. This work introduces a relational learning approach that models inter-class structure using dynamic class prototypes and a sparsified similarity graph with graph-based refinement. Experiments on CUB-200-2011, FGVC-Aircraft, and Stanford Cars demonstrate consistent improvements over DTRG. Our model achieves 2.35% Top-1 improvement on Aircraft, 1.34% on CUB, and 2.29% on Cars, while also improving Top-5 accuracy and F1-score across datasets. These results demonstrate that relational modeling of evolving class representations improves fine-grained recognition.