SOC: Semantic-Assisted Object Cluster for Referring Video Object Segmentation

Benchmark Model Rank Results
referring-expression-segmentation-on-a2d-sentencesSOC (Video-Swin-B)#2AP: 0.573IoU overall: 0.807IoU mean: 0.725Precision@0.5: 0.851…
referring-expression-segmentation-on-a2d-sentencesSOC (Video-Swin-T)#4AP: 0.504IoU overall: 0.747IoU mean: 0.669Precision@0.5: 0.79…
referring-expression-segmentation-on-j-hmdbSOC (Video-Swin-B)#2AP: 0.446IoU overall: 0.736IoU mean: 0.723Precision@0.5: 0.969…
referring-expression-segmentation-on-j-hmdbSOC (Video-Swin-T)#3AP: 0.397IoU overall: 0.707IoU mean: 0.701Precision@0.5: 0.947…
referring-video-object-segmentation-on-long-rvosSOC#6J&F: 34.9tIoU: 68.1vIoU: 28.6
referring-video-object-segmentation-on-refer-youtube-vosSOC#7J&F: 66.0J: 64.1F: 67.9
referring-expression-segmentation-on-refer-youtube-vos-2021-public-validationSOC (Joint training, Video-Swin-B)#19J&F: 67.3±0.5J: 65.3F: 69.3
referring-expression-segmentation-on-refer-youtube-vos-2021-public-validationSOC (Video-Swin-T)#37J&F: 59.2J: 57.8F: 60.5
referring-video-object-segmentation-on-ref-davis17SOC#28J&F: 65.8J: 62.5F: 69.1