• Add SwinV2, FocalNet, GBR COTS dataset
• Update codes for mmdet 2.25.0, mmcv-full 1.4.4

• Support finer scale-wise AP metrics
• Add codes for TOOD, ConvMLP, PoolFormer
• Update codes for PyTorch 1.9.0, mmdet 2.17.0, mmcv-full 1.3.13

This release includes codes for SOTA methods.
See here for the weights of pre-trained models after this release.

This release provides COCO pre-trained UniverseNet 20.08s models and Manga109-s pre-trained models.

This release provides COCO pre-trained models of UniverseNet 20.08 variants.

This release provides COCO pre-trained models of UniverseNet 20.08.

This release provides COCO pre-trained models of UniverseNet+GFL.
The version of MMDetection is updated to 2.3.0rc0. CUDA ops are moved to MMCV.

This release provides COCO pre-trained models of UniverseNet.

We also attached an ImageNet pre-trained Res2Net-v1b-50 converted from https://github.com/Res2Net/Res2Net-PretrainedModels .
To reduce GPU memory consumption, please use this weight file.