Video Semantic Segmentation on VSPW (50% warm-up ratio W2F protocol)
52.3mIoUDiTTA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DiTTAApproach=ISS+SAM2, FPS=13.452026.04 | 52.3 | 67.1 | 94.9 | 93 | |
| Zero-shot Refine.Approach=ISS+SAM2, FPS=1.412026.04 | 50.1 | 67.3 | 95 | 93.3 | |
| TV3SApproach=VSS, FPS=5.712026.04 | 49.8 | 66.7 | 91.3 | 87.6 | |
| CoTTAApproach=ISS+TTA, FPS=18.482026.04 | 49.8 | 66.7 | 90.6 | 87.5 | |
| CFFMApproach=VSS, FPS=5.982026.04 | 49.7 | 66.5 | 90.8 | 87.2 | |
| CFFM++Approach=VSS, FPS=5.852026.04 | 49.7 | 66.2 | 90.6 | 86.9 | |
| SegFormerApproach=ISS, FPS=18.582026.04 | 48.7 | 66.4 | 88.3 | 84.2 | |
| AuxAdaptApproach=ISS+TTA, FPS=18.562026.04 | 48.7 | 65.6 | 90.5 | 87.3 | |
| TENTApproach=ISS+TTA, FPS=18.452026.04 | 47.6 | 64.3 | 92.6 | 90.3 |