Enhancing fine-grained image classification through attentive batch training

Duy M. Le, Bao Q. Bui, Anh Tran, Cong Tran, Cuong Pham

Abstract: Fine-grained image classification, which is a challenging task in computer vision, requires precise differentiation among visually similar object categories. In this paper, we design a novel framework, namely Relationship Batch Integration (RBI), allowing the discernment of vital visual features that may remain elusive when examining a singular image representative of a particular class. Our proposed method, validated through extensive experiments, significantly boosts the accuracy of fine-grained classifiers, achieving state-of-the-art performance with (97.79%)<math><mrow is="true"><mo is="true">(</mo><mn is="true">97</mn><mo is="true">.</mo><mn is="true">79</mn><mtext is="true">%</mtext><mo is="true">)</mo></mrow></math> on the Stanford Dog dataset, even attaining a top result of (93.71%)<math><mrow is="true"><mo is="true">(</mo><mn is="true">93</mn><mo is="true">.</mo><mn is="true">71</mn><mtext is="true">%</mtext><mo is="true">)</mo></mrow></math> on the Tiny-Imagenet dataset for general image classification.

External IDs:doi:10.1016/j.icte.2025.10.002