DBN-Mix: Training dual branch network using bilateral mixup augmentation for long-tailed visual recognition

Jae Soon Baik, In Young Yoon, Jun Won Choi

Abstract: Highlights•The proposed bilateral mixup improves representation learning for minority class.•Class-wise temperature scaling effectively mitigates the bias in classifier.•Dual-branch network with two main ideas achieves state-of-the-art performance.•Unlike previous two-stage methods, DBN-Mix enables end-to-end training.•Simple implementation with only a few lines of code.