Video label propagation on JHMDB (val)
63.4PCK@0.1DIFTadm
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DIFTadmpre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 63.4 | 84.3 | |
| SFCpre-trained on videos=true, pre-training dataset=YT-VOS2023.06 | 61.9 | 83 | |
| DIFTsdpre-trained on videos=false, pre-training dataset=LAION2023.06 | 61.1 | 81.8 | |
| VFSpre-trained on videos=true, pre-training dataset=Kinetic2023.06 | 60.9 | 80.7 | |
| SimSiampre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 59.9 | 81.6 | |
| MoCopre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 59.4 | 80.9 | |
| SimCLRpre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 59 | 80.8 | |
| BYOLpre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 58.8 | 80.9 | |
| VINCEpre-trained on videos=true, pre-training dataset=Kinetic2023.06 | 58.8 | 80.4 | |
| CRWpre-trained on videos=true, pre-training dataset=Kinetic2023.06 | 58.8 | 80.3 | |
| UVCpre-trained on videos=true, pre-training dataset=Kinetic2023.06 | 58.6 | 79.6 | |
| InstDispre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 58.5 | 80.2 | |
| CorrFlowpre-trained on videos=true, pre-training dataset=OxUvA2023.06 | 58.5 | 78.8 | |
| TimeCyclepre-trained on videos=true, pre-training dataset=VLOG2023.06 | 57.3 | 78.1 | |
| DINOpre-trained on videos=false, pre-training dataset=ImageNet w/o labels2023.06 | 57.2 | 81.2 | |
| Colorizationpre-trained on videos=true, pre-training dataset=Kinetic2023.06 | 45.2 | 69.6 | |
| OpenCLIPpre-trained on videos=false, pre-training dataset=LAION2023.06 | 41.7 | 71.7 |