Title:Seesaw Loss for Long-Tailed Instance Segmentation

Abstract:This report presents the approach used in the submission of the LVIS Challenge 2020 of team MMDet. In the submission, we propose Seesaw Loss that dynamically rebalances the penalty to each category according to a relative ratio of cumulative training instances between different categories. Furthermore, we propose HTC-Lite, a light-weight version of Hybrid Task Cascade (HTC) which replaces the semantic segmentation branch by a global context encoder. Seesaw Loss improves the strong baseline by 6.9% AP on LVIS v1 val split. With a single model, and without using external data and annotations except for standard ImageNet-1k classification dataset for backbone pre-training, our submission achieves 38.92% AP on the test-dev split of the LVIS v1 benchmark.