Video Semantic Segmentation on VSPW (val)
65.7mIoUPMT
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| PMTBackbone=ViT-L, Pre-training=DINOv3, Encoder=frozen, GFLOPs=2049, FPS=572026.03 | 65.7 | — | — | 94.9 | — | — | — | |
| VidEoMTBackbone=ViT-L, Pre-training=DINOv2, GFLOPs=1909, FPS=732026.02 | 64.9 | — | — | 95 | — | — | — | |
| VidEoMTBackbone=ViT-L, Pre-training=DINOv2, Encoder=finetuned, GFLOPs=1909, FPS=732026.03 | 64.9 | — | — | 95 | — | — | — | |
| PMTBackbone=ViT-L, Pre-training=DINOv2, Encoder=frozen, GFLOPs=2049, FPS=602026.03 | 64.3 | — | — | 94.6 | — | — | — | |
| VidEoMTBackbone=ViT-L, Pre-training=DINOv3, Encoder=finetuned, GFLOPs=1909, FPS=712026.03 | 64 | — | — | 94.4 | — | — | — | |
| DVIS++Evaluation Protocol=Supervised Segmentation, Backbone=VIT-L2026.03 | 63.8 | — | — | 95.1 | 95.7 | — | — | |
| TubeFormerBackbone=Axial-ResNet50x642023.03 | 63.2 | — | — | 88 | 92.1 | — | — | |
| TubeFormer-DeepLab-B4Backbone=B42022.05 | 63.16 | — | — | 87.95 | 92.08 | — | — | |
| DVIS++Backbone=ViT-L, Pre-training=DINOv2, GFLOPs=2290, FPS=132026.02 | 62.8 | — | — | 94.2 | — | — | — | |
| DVIS++Backbone=ViT-Adapter-L, Pre-training=DINOv2, Encoder=finetuned, GFLOPs=2290, FPS=132026.03 | 62.8 | — | — | 94.2 | — | — | — | |
| CharlesBLWX2022.05 | 61.44 | — | — | — | — | — | — | |
| DVISBackbone=Swin-L, Pre-training=IN21K, GFLOPs=879, FPS=222026.02 | 61.3 | — | — | 94.3 | — | — | — | |
| DVISBackbone=Swin-L, Pre-training=ImageNet-21K, Encoder=finetuned, GFLOPs=879, FPS=222026.03 | 61.3 | — | — | 94.3 | — | — | — | |
| VidSegDiffEvaluation Protocol=Zero-Shot Segmentation, Backbone=SD 2.12026.03 | 60.6 | — | — | 89.6 | 90.7 | — | — | |
| UniVSEvaluation Protocol=Supervised Segmentation, Backbone=Swin-T2026.03 | 59.8 | — | — | — | 92.3 | — | — | |
| Tube-LinkBackbone=Swin-large2023.03 | 59.7 | — | — | 88.4 | 90.3 | — | — | |
| jjRain2022.05 | 59.3 | — | — | 86.87 | 90.07 | — | — | |
| BetterThing2022.05 | 57.89 | — | — | — | — | — | — | |
| VidSegDiffEvaluation Protocol=Zero-Shot Segmentation, Backbone=SVD2026.03 | 53.2 | — | — | 88 | 89.3 | — | — | |
| DiTTAApproach=ISS + SAM2, Adaptation Protocol=Full-video2026.04 | 53.2 | 68.1 | — | 94.3 | 95.9 | — | — | |
| THE-MaskApproach=VSS, Adaptation Protocol=Full-video2026.04 | 52.1 | 67.2 | — | — | — | — | — | |
| TV3SBackbone=Swin-S, GFLOPs=94.1, Params(M)=53.1, FPS=19.5, Input resolution=480x8532025.03 | 50.6 | — | — | 85.8 | 89.6 | — | — | |
| CFFM++Backbone=MiT-B5, Params (M)=87.5, Input Size=480x8532022.04 | 50.1 | 66.5 | — | 87.4 | 90.8 | 249.5 | 10.4 | |
| CFFM++Approach=VSS, Adaptation Protocol=Full-video2026.04 | 50.1 | 66.5 | — | 87.4 | 90.8 | — | — | |
| AuxAdapt*Approach=ISS+TTA, Adaptation Protocol=Full-video2026.04 | 50 | 66.8 | — | 89.2 | 92.4 | — | — | |
| MRCFABackbone=MiT-B5, Params (M)=84.5, FPS (f/s)=15.72022.07 | 49.9 | 66 | 90.9 | 87.4 | — | — | — | |
| MRCFABackbone=MiT-B52023.03 | 49.9 | — | — | 87.4 | 90.9 | — | — | |
| MRCFABackbone=MiT-B5, GFLOPs=373, Params(M)=84.5, FPS=5, Input resolution=480x8532025.03 | 49.9 | — | — | 87.4 | 90.9 | — | — | |
| TV3SBackbone=MiT-B5, GFLOPs=137, Params(M)=85.6, FPS=14, Input resolution=480x8532025.03 | 49.8 | — | — | 88.7 | 91.7 | — | — | |
| TV3SApproach=VSS, Adaptation Protocol=Full-video2026.04 | 49.8 | — | — | 88.7 | 91.7 | — | — | |
| CoTTA*Approach=ISS+TTA, Adaptation Protocol=Full-video2026.04 | 49.4 | 65.9 | — | 83.3 | 87.4 | — | — | |
| CFFMBackbone=MiT-B52023.03 | 49.3 | — | — | 87.1 | 90.8 | — | — | |
| CFFMBackbone=MiT-B5, Params (M)=85.5, Input Size=480x8532022.04 | 49.3 | 65.8 | — | 87.1 | 90.8 | 232.2 | 11.3 | |
| CFFMBackbone=MiT-B5, GFLOPs=413.5, Params(M)=85.5, FPS=4.5, Input resolution=480x8532025.03 | 49.3 | — | — | 87.1 | 90.8 | — | — | |
| CFFMApproach=VSS, Adaptation Protocol=Full-video2026.04 | 49.3 | 65.8 | — | 87.1 | 90.8 | — | — | |
| TENT*Approach=ISS+TTA, Adaptation Protocol=Full-video2026.04 | 49.3 | 65.9 | — | 83.3 | 87.4 | — | — | |
| Zero-shot Refine.Approach=ISS + SAM2, Adaptation Protocol=Full-video2026.04 | 49.2 | 66.5 | — | 93 | 94.7 | — | — | |
| SegFormerApproach=ISS, Adaptation Protocol=Full-video2026.04 | 49 | 66.3 | — | 84.2 | 88.2 | — | — | |
| SegformerBackbone=MiT-B5, GFLOPs=185, Params(M)=82.1, FPS=9.4, Input resolution=480x8532025.03 | 48.9 | — | — | 83.7 | 87.8 | — | — | |
| SegFormerBackbone=MiT-B5, Params (M)=82.1, FPS (f/s)=17.22022.07 | 48.2 | 65.1 | 87.8 | 83.7 | — | — | — | |
| SegFormerBackbone=MiT-B5, Params (M)=82.1, Input Size=480x8532022.04 | 48.2 | 65.1 | — | 83.7 | 87.8 | 187 | 17.2 | |
| TV3SBackbone=MiT-B2, GFLOPs=53.9, Params(M)=28.3, FPS=21.9, Input resolution=480x8532025.03 | 46.3 | — | — | 88.35 | 91.5 | — | — | |
| CFFM++Backbone=MiT-B2, Params (M)=28.5, Input Size=480x8532022.04 | 45.5 | 64.7 | — | 86.4 | 90.2 | 96.9 | 21.5 | |
| MRCFABackbone=MiT-B2, Params (M)=27.3, FPS (f/s)=32.12022.07 | 45.3 | 64.7 | 90.3 | 86.2 | — | — | — | |
| MRCFABackbone=MiT-B2, GFLOPs=127.9, Params(M)=27.3, FPS=10.7, Input resolution=480x8532025.03 | 45.3 | — | — | 86.2 | 90.3 | — | — | |
| CFFMBackbone=MiT-B2, Params (M)=26.5, Input Size=480x8532022.04 | 44.9 | 64.9 | — | 85.8 | 89.8 | 79.6 | 23.8 | |
| CFFMBackbone=MiT-B2, GFLOPs=143.2, Params(M)=26.5, FPS=10.1, Input resolution=480x8532025.03 | 44.9 | — | — | 85.8 | 89.8 | — | — | |
| TV3SBackbone=Swin-T, GFLOPs=57.3, Params(M)=31.7, FPS=22.9, Input resolution=480x8532025.03 | 44.9 | — | — | 83.5 | 88 | — | — | |
| SegFormerBackbone=MiT-B2, Params (M)=24.8, FPS (f/s)=39.22022.07 | 43.9 | 63.7 | 86 | 81.2 | — | — | — | |
| SegFormerBackbone=MiT-B2, Params (M)=24.8, Input Size=480x8532022.04 | 43.9 | 63.7 | — | 81.2 | 86 | 57.2 | 39.2 | |
| SegformerBackbone=MiT-B2, GFLOPs=100.8, Params(M)=24.8, FPS=16.2, Input resolution=480x8532025.03 | 43.9 | — | — | 81.2 | 86 | — | — | |
| EmerDiffEvaluation Protocol=Zero-Shot Segmentation, Backbone=SD 2.12026.03 | 43.4 | — | — | 64.3 | 68.9 | — | — | |
| Tube-LinkBackbone=ResNet502023.03 | 42.3 | — | — | 83.2 | 86.8 | — | — | |
| Mask2FormerBackbone=Swin-S, Frame-by-Frame processing=true, GFLOPs=152.2, Params(M)=68.9, FPS=14.5, Input resolution=480x8532025.03 | 42.1 | — | — | 79.3 | 84.7 | — | — | |
| Mask2FormerBackbone=Swin-T, GFLOPs=114.4, Params(M)=47.4, FPS=17.1, Input resolution=480x8532025.03 | 41.2 | — | — | 80 | 84.5 | — | — | |
| MPVSSBackbone=Swin-S, GFLOPs=47.3, Params(M)=108, FPS=30.6, Input resolution=480x8532025.03 | 40.4 | — | — | 80.7 | 86 | — | — | |
| TV3SBackbone=MiT-B1, GFLOPs=36.9, Params(M)=17.3, FPS=24.7, Input resolution=480x8532025.03 | 40 | — | — | 87 | 90.7 | — | — | |
| CFFM++Backbone=MiT-B1, Params (M)=16.5, Input Size=480x8532022.04 | 39.9 | 60.7 | — | 84.9 | 89.1 | 64.2 | 27.6 | |
| MPVSSBackbone=Swin-T, GFLOPs=39.7, Params(M)=114, FPS=32.8, Input resolution=480x8532025.03 | 39.9 | — | — | 80.4 | 85.9 | — | — | |
| EmerDiffEvaluation Protocol=Zero-Shot Segmentation, Backbone=SVD2026.03 | 39.7 | — | — | 78.5 | 82.1 | — | — | |
| Mask2FormerBackbone=R101, Frame-by-Frame processing=true, GFLOPs=141.3, Params(M)=63, FPS=16.9, Input resolution=480x8532025.03 | 39.3 | — | — | 77.6 | 82.5 | — | — | |
| MRCFABackbone=MiT-B1, Params (M)=16.2, FPS (f/s)=40.12022.07 | 38.9 | 60 | 88.8 | 84.4 | — | — | — | |
| MRCFABackbone=MiT-B1, Params(M)=16.2, FPS=40.1, Input resolution=480x8532025.03 | 38.9 | — | — | 84.4 | 88.8 | — | — | |
| MPVSSBackbone=R101, GFLOPs=45.1, Params(M)=103.1, FPS=32.3, Input resolution=480x8532025.03 | 38.8 | — | — | 79.6 | 84.8 | — | — | |
| CFFMBackbone=MiT-B1, Params (M)=15.5, Input Size=480x8532022.04 | 38.5 | 60 | — | 84.1 | 88.6 | 55.4 | 29.8 | |
| CFFMBackbone=MiT-B1, Params(M)=15.5, FPS=29.8, Input resolution=480x8532025.03 | 38.5 | — | — | 84.1 | 88.6 | — | — | |
| Mask2FormerBackbone=R50, Frame-by-Frame processing=true, GFLOPs=110.6, Params(M)=44, FPS=19.4, Input resolution=480x8532025.03 | 38.5 | — | — | 76.4 | 81.3 | — | — | |
| Video K-Net (PSPNet)Backbone=ResNet1012022.04 | 38 | — | — | 82.3 | 87.2 | — | — | |
| Video K-Net (PSPNet)Backbone=ResNet1012023.03 | 38 | — | — | 82.3 | 87.2 | — | — | |
| Video K-Net (Deeplabv3+)Backbone=ResNet1012022.04 | 37.9 | — | — | 82.1 | 87 | — | — | |
| Video K-Net (Deeplabv3+)Backbone=ResNet1012023.03 | 37.9 | — | — | 82.1 | 87 | — | — | |
| TCB2022.05 | 37.82 | — | — | 83.99 | 87.86 | — | — | |
| TCBst-ocr-memBackbone=ResNet-101, Params (M)=58.1, FPS (f/s)=5.52022.07 | 37.8 | 59.5 | 87.9 | 84 | — | — | — | |
| TCBst-ocr-memBackbone=ResNet-101, Params (M)=58.1, Input Size=480x8532022.04 | 37.8 | 59.5 | — | 84 | 87.9 | — | 5.5 | |
| TCBBackbone=R101, Frame-by-Frame processing=true, GFLOPs=1692, Input resolution=480x8532025.03 | 37.8 | — | — | 84 | 87.9 | — | — | |
| ETCBackbone=OCRNet, Params (M)=58.12022.07 | 37.5 | 59.1 | 84.1 | 79.1 | — | — | — | |
| NetWarpBackbone=OCRNet, Params (M)=58.12022.07 | 37.5 | 58.9 | 84 | 79 | — | — | — | |
| TCBst-ppmBackbone=ResNet-101, Params (M)=70.5, FPS (f/s)=10.02022.07 | 37.5 | 58.6 | 87 | 82.1 | — | — | — | |
| TCB(PSPNet)Backbone=ResNet1012022.04 | 37.5 | — | — | 82.1 | 86.9 | — | — | |
| TCB(PSPNet)Backbone=ResNet1012023.03 | 37.5 | — | — | 82.1 | 86.9 | — | — | |
| ETCBackbone=OCRNet, Params (M)=58.1, Input Size=480x8532022.04 | 37.5 | 59.1 | — | 79.1 | 84.1 | — | — | |
| NetWarpBackbone=OCRNet, Params (M)=58.1, Input Size=480x8532022.04 | 37.5 | 58.9 | — | 79 | 84 | — | — | |
| TCBst-ppmBackbone=ResNet-101, Params (M)=70.5, Input Size=480x8532022.04 | 37.5 | 58.6 | — | 82.1 | 87 | — | 10 | |
| MPVSSBackbone=R50, GFLOPs=38.9, Params(M)=84.1, FPS=33.9, Input resolution=480x8532025.03 | 37.5 | — | — | 77.2 | 84.1 | — | — | |
| ETCBackbone=OCRNet, Frame-by-Frame processing=true, GFLOPs=361.7, Input resolution=480x8532025.03 | 37.5 | — | — | 79.1 | 84.1 | — | — | |
| TCBst-ocrBackbone=ResNet-101, Params (M)=58.1, FPS (f/s)=5.52022.07 | 37.4 | 59.3 | 86.9 | 82 | — | — | — | |
| TCBst-ocrBackbone=ResNet-101, Params (M)=58.1, Input Size=480x8532022.04 | 37.4 | 59.3 | — | 82 | 86.9 | — | 5.5 | |
| NetWarpBackbone=PSPNet, Params (M)=89.42022.07 | 37 | 57.9 | 84.4 | 79.4 | — | — | — | |
| NetWarpBackbone=PSPNet, Params (M)=89.4, Input Size=480x8532022.04 | 37 | 57.9 | — | 79.4 | 84.4 | — | — | |
| OCRNetBackbone=ResNet-101, Params (M)=58.1, FPS (f/s)=14.32022.07 | 36.7 | 59.2 | 84 | 79 | — | — | — | |
| OCRNetBackbone=ResNet-101, Params (M)=58.1, Input Size=480x8532022.04 | 36.7 | 59.2 | — | 79 | 84 | — | 14.3 | |
| OCRNetBackbone=R101, Frame-by-Frame processing=true, GFLOPs=361.7, Params(M)=58.1, FPS=14.3, Input resolution=480x8532025.03 | 36.7 | — | — | 79 | 84 | — | — | |
| ETCBackbone=PSPNet, Params (M)=89.42022.07 | 36.6 | 58.3 | 84.1 | 79.2 | — | — | — | |
| ETCBackbone=PSPNet, Params (M)=89.4, Input Size=480x8532022.04 | 36.6 | 58.3 | — | 79.2 | 84.1 | — | — | |
| SegFormerBackbone=MiT-B1, Params (M)=13.8, FPS (f/s)=58.72022.07 | 36.5 | 58.8 | 84.7 | 79.9 | — | — | — | |
| UperNetBackbone=ResNet-101, Params (M)=83.22022.07 | 36.5 | 58.6 | 82.6 | 76.1 | — | — | — | |
| PSPNetBackbone=ResNet-101, Params (M)=70.5, FPS (f/s)=13.92022.07 | 36.5 | 58.1 | 84.2 | 79.6 | — | — | — | |
| PSPNet+Backbone=ResNet1012022.04 | 36.5 | — | — | 79.8 | 84.4 | — | — | |
| SegFormerBackbone=MiT-B1, Params (M)=13.8, Input Size=480x8532022.04 | 36.5 | 58.8 | — | 79.9 | 84.7 | 33 | 58.7 | |
| UperNetBackbone=ResNet-101, Params (M)=83.2, Input Size=480x8532022.04 | 36.5 | 58.6 | — | 76.1 | 82.6 | — | — |