DeVIS: Making Deformable Transformers Work for Video Instance Segmentation

Benchmark Model Rank Results
video-instance-segmentation-on-ovis-1DeVIS (Swin-L)#22mask AP: 35.5AP50: 59.3AP75: 38.3AR1: 16.6AR10: 39.8
video-instance-segmentation-on-ovis-1DeVIS (ResNet-50)#30mask AP: 23.7AP50: 47.6AP75: 20.8AR1: 12.0AR10: 28.9
video-instance-segmentation-on-youtube-vis-1DeVIS (Swin-L)#10mask AP: 57.1AP50: 80.8AP75: 66.3AR1: 50.8AR10: 61.0
video-instance-segmentation-on-youtube-vis-1DeVIS (ResNet-50)#21mask AP: 44.4AP50: 66.7AP75: 48.6AR1: 42.4AR10: 51.6
video-instance-segmentation-on-youtube-vis-2DeVIS (Swin-L)#15mask AP: 54.4AP50: 77.7AP75: 59.8AR1: 43.8AR10: 57.8
video-instance-segmentation-on-youtube-vis-2DeVIS (ResNet-50)#21mask AP: 43.1AP50: 66.8AP75: 46.6AR1: 38.0AR10: 50.1