Asymmetric Loss For Multi-Label Classification
This addresses the challenge of imbalanced label distributions in multi-label classification for computer vision applications, offering an incremental improvement over existing methods.
The paper tackles the problem of positive-negative imbalance in multi-label classification, which can lead to poor accuracy by under-emphasizing gradients from positive labels. It introduces an asymmetric loss (ASL) that dynamically down-weights and hard-thresholds easy negative samples, achieving state-of-the-art results on datasets like MS-COCO, Pascal-VOC, NUS-WIDE, and Open Images.
In a typical multi-label setting, a picture contains on average few positive labels, and many negative ones. This positive-negative imbalance dominates the optimization process, and can lead to under-emphasizing gradients from positive labels during training, resulting in poor accuracy. In this paper, we introduce a novel asymmetric loss ("ASL"), which operates differently on positive and negative samples. The loss enables to dynamically down-weights and hard-thresholds easy negative samples, while also discarding possibly mislabeled samples. We demonstrate how ASL can balance the probabilities of different samples, and how this balancing is translated to better mAP scores. With ASL, we reach state-of-the-art results on multiple popular multi-label datasets: MS-COCO, Pascal-VOC, NUS-WIDE and Open Images. We also demonstrate ASL applicability for other tasks, such as single-label classification and object detection. ASL is effective, easy to implement, and does not increase the training time or complexity. Implementation is available at: https://github.com/Alibaba-MIIL/ASL.