The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. While training with class-balanced sampling has been shown effective for this problem, it is known to over-fit to few-shot classes. It is hypothesized that this is due to the repeated sampling of examples and can be addressed by feature space augmentation. A new feature augmentation strategy, EMANATE, based on back-tracking of features across epochs during training, is proposed. It is shown that, unlike class-balanced sampling, this is an adversarial augmentation strategy. A new sampling procedure, Breadcrumb, is then introduced to implement adversarial class-balanced sampling without extra computation. Experiments on three popular long-tailed recognition datasets show that Breadcrumb training produces classifiers that outperform existing solutions to the problem.
@article{arxiv.2105.00127,
title = {Breadcrumbs: Adversarial Class-Balanced Sampling for Long-tailed Recognition},
author = {Bo Liu and Haoxiang Li and Hao Kang and Gang Hua and Nuno Vasconcelos},
journal= {arXiv preprint arXiv:2105.00127},
year = {2021}
}