3D Human Mesh Recovery on BEDLAM
36.9PA-PVEMulti-HMR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Multi-HMREncoder Arch.=ViT-Large, Training Data=HMR-500K, Training Res.=6722025.03 | 36.9 | 95 | — | |
| DUNEEncoder Arch.=ViT-Base, Training Data=DUNE-20.7M, Training Res.=4482025.03 | 56 | 94 | — | |
| STARHyperparameter Selection Strategy=TAP2026.04 | 59.3 | — | 96.2 | |
| TAHyperparameter Selection Strategy=TAP2026.04 | 59.7 | — | 95.6 | |
| DUNEStatus=Pretrained (reproduction)2026.04 | 62.5 | — | 92.9 | |
| AdaMergingHyperparameter Selection Strategy=TAP2026.04 | 62.9 | — | 97.1 | |
| TSVHyperparameter Selection Strategy=TAP2026.04 | 62.9 | — | 93.6 | |
| TIESHyperparameter Selection Strategy=TAP2026.04 | 63.3 | — | 93.4 | |
| ConsensusHyperparameter Selection Strategy=TAP2026.04 | 65.7 | — | 92.8 | |
| DUNEEncoder Arch.=ViT-Base, Training Data=DUNE-20.7M, Training Res.=3362025.03 | 68.3 | 91 | — | |
| DINO-v2Encoder Arch.=ViT-Base, Training Data=LVD-142M, Training Res.=5182025.03 | 76.5 | 86 | — | |
| AM-RADIO-v2.5Encoder Arch.=ViT-Base, Training Data=DataComp-1B, Training Res.=5122025.03 | 83.2 | 89 | — |