UniVS: Unified and Universal Video Segmentation with Prompts as Queries

Benchmark Model Rank Results
video-object-segmentation-on-youtube-vos-2018UniVS(Swin-L)#2Mean Jaccard & F-Measure: 71.5
video-object-segmentation-on-davis-2017-valUniVS(Swin-L)#3Mean Jaccard & F-Measure: 76.2F-measure: 79.5Jaccard: 72.8
video-semantic-segmentation-on-vspwUniVS(Swin-L)#5mIoU: 59.8
referring-expression-segmentation-on-davis-2017-valUniVS(Swin-L)#8J&F 1st frame: 59.4?J&F Full video: 59.4
video-instance-segmentation-on-youtube-vis-validationUniVS(Swin-L)#11mask AP: 60.0AP50: 82.1AP75: 65.3AR1: 54.7AR10: 66.8
video-panoptic-segmentation-on-vipsegUniVS(Swin-L)#13VPQ: 49.3STQ: 58.2
video-instance-segmentation-on-youtube-vis-2021UniVS(Swin-L)#14mask AP: 57.9AP50: 79.4AP75: 63.3AR1: 46.2AR10: 63.1
video-instance-segmentation-on-ovis-validationUniVS(Swin-L)#20mask AP: 41.7
referring-expression-segmentation-on-refer-youtube-vos-2021-public-validationUniVS(Swin-L)#38J&F: 58.0J: 56.8F: 59.5