Text-to-motion generation on HumanML3D full dimension (test)
52.3R-Precision Top 1MARDM
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| MARDM2024.11 | 52.3 | 71.5 | 81 | 0.061 | — | — | — | |
| MoMaskReproduced=false2024.11 | 52.1 | 71.3 | 80.7 | 0.045 | — | — | — | |
| MMMReproduced=true2024.11 | 51.7 | 71.1 | 80.5 | 0.071 | — | — | — | |
| MMMReproduced=false2024.11 | 51.5 | 70.8 | 80.4 | 0.089 | — | — | — | |
| GTtype=Ground Truth2024.11 | 51.1 | 70.3 | 79.7 | 0.002 | — | — | — | |
| MoMaskReproduced=true2024.11 | 50.8 | 70.1 | 79.6 | 0.093 | — | — | — | |
| GT Joints HumanML3Dtype=Ground Truth (Joints)2024.11 | 50.3 | 69.7 | 78.9 | 0.015 | — | — | — | |
| Ours-SiTFramework=Diffusion2024.11 | 50 | 69.5 | 79.5 | 0.114 | 3.27 | 2.231 | 0.642 | |
| T2M-GPTReproduced=true2024.11 | 49.7 | 68.5 | 77.9 | 0.115 | — | — | — | |
| Ours-DDPMFramework=Autoregressive2024.11 | 49.2 | 69 | 79 | 0.116 | 3.349 | 2.47 | 0.637 | |
| T2M-GPTReproduced=false2024.11 | 49.2 | 67.9 | 77.5 | 0.141 | — | — | — | |
| MoMaskFramework=VQ2024.11 | 49 | 68.7 | 78.6 | 0.116 | 3.353 | 1.263 | 0.637 | |
| MMMFramework=VQ2024.11 | 48.7 | 68.3 | 78.2 | 0.132 | 3.359 | 1.241 | 0.635 | |
| T2M-GPTFramework=VQ2024.11 | 47 | 65.9 | 75.8 | 0.335 | 3.505 | 2.018 | 0.607 | |
| ReMoDiffuseFramework=Diffusion2024.11 | 46.8 | 65.3 | 75.4 | 0.883 | 3.414 | 2.703 | 0.621 | |
| MLDFramework=Diffusion2024.11 | 46.1 | 65.1 | 75 | 0.431 | 3.445 | 3.506 | 0.61 | |
| MDMSteps=50, Reproduced=true2024.11 | 45.9 | 65.1 | 75.3 | 0.481 | — | — | — | |
| MDMSteps=502024.11 | 45.5 | 64.5 | 74.9 | 0.489 | — | — | — | |
| MotionDiffuseFramework=Diffusion2024.11 | 45 | 64.1 | 75.3 | 0.778 | 3.49 | 3.179 | 0.606 | |
| MDM-50StepFramework=Diffusion2024.11 | 44 | 63.6 | 74.2 | 0.518 | 3.64 | 3.604 | 0.578 |