Equalization Loss v2: A New Gradient Balance Approach for Long-tailed Object Detection
This work provides an incremental improvement in end-to-end training methods for long-tailed object detection, benefiting practitioners working with imbalanced datasets.
The paper addresses the issue of imbalanced gradients between positive and negative samples in long-tailed object detection. The proposed Equalization Loss v2 (EQL v2) re-balances the training process for each category, outperforming the original EQL by 4 AP overall and 14-18 AP on rare categories on the LVIS benchmark, and also surpassing decoupled training methods.
Recently proposed decoupled training methods emerge as a dominant paradigm for long-tailed object detection. But they require an extra fine-tuning stage, and the disjointed optimization of representation and classifier might lead to suboptimal results. However, end-to-end training methods, like equalization loss (EQL), still perform worse than decoupled training methods. In this paper, we reveal the main issue in long-tailed object detection is the imbalanced gradients between positives and negatives, and find that EQL does not solve it well. To address the problem of imbalanced gradients, we introduce a new version of equalization loss, called equalization loss v2 (EQL v2), a novel gradient guided reweighing mechanism that re-balances the training process for each category independently and equally. Extensive experiments are performed on the challenging LVIS benchmark. EQL v2 outperforms origin EQL by about 4 points overall AP with 14-18 points improvements on the rare categories. More importantly, it also surpasses decoupled training methods. Without further tuning for the Open Images dataset, EQL v2 improves EQL by 7.3 points AP, showing strong generalization ability. Codes have been released at https://github.com/tztztztztz/eqlv2