This model introduces a novel framework, Relationship Batch Integration (RBI), designed to enhance performance on challenging fine-grained image classification tasks. Instead of analyzing images in isolation, the model leverages the relationships between all images within a training batch to identify subtle, distinguishing visual features that are crucial for differentiating between visually similar categories.

@article{le2024enhancing,
  title={Enhancing Fine-grained Image Classification through Attentive Batch Training},
  author={Le, Duy M and Bui, Bao Q and Tran, Anh and Tran, Cong and Pham, Cuong},
  journal={arXiv preprint arXiv:2412.19606},
  year={2024}
}