Research map / Video / Video generation
Human Joint Kinematics Diffusion-Refinement for Stochastic Motion Prediction
Dong Wei, Huaijiang Sun, Bin Li, Jianfeng Lu, Weiqing Li, Xiaoning Sun, Shengxiang Hu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use diffusion-refinement instead of DLow | stochastic human motion prediction · Human3.6M | APD: 11.741 -> 15.353 | Improved |
| use diffusion-refinement instead of DLow | stochastic human motion prediction · HumanEva-I | APD: 4.855 -> 5.931 | Improved |
| add the GCN refinement network to diffusion (ablation) | MotionDiff · Human3.6M | ADE: 0.486 -> 0.411 | Improved |
| add the GCN refinement network to diffusion (ablation) | MotionDiff · HumanEva-I | ADE: 0.273 -> 0.232 | Improved |
| add the GCN refinement network to diffusion (ablation) | MotionDiff · Human3.6M | APD: 15.534 -> 15.353 | Hurt |
| add the GCN refinement network to diffusion (ablation) | MotionDiff · HumanEva-I | APD: 6.508 -> 5.931 | Hurt |
| use joint embedding and spatial transformer instead of linear pose projection (ablation) | MotionDiff · Human3.6M | FDE: 0.535 -> 0.509 | Improved |
| replace temporal transformer with MLP (ablation) | MotionDiff · Human3.6M | APD: 15.353 -> 10.943 | Hurt |
| set spatial transformer dimension to 48 instead of 32 (ablation) | MotionDiff · Human3.6M | APD: 15.353 -> 15.761 | Improved |
| set spatial transformer dimension to 48 instead of 32 (ablation) | MotionDiff · Human3.6M | FDE: 0.509 -> 0.543 | Hurt |
| set temporal transformer dimension to 256 instead of 512 (ablation) | MotionDiff · Human3.6M | APD: 15.353 -> 12.4735 | Hurt |
| use L1 instead of L2 distance (ablation) | MotionDiff · Human3.6M | APD: 15.353 -> 13.894 | Hurt |
Closest papers by meaning
- KinemaDiff: Towards Diffusion for Coherent and Physically Plausible Human Motion Prediction (ICLR 2026 · similarity 0.92)
- Nonisotropic Gaussian Diffusion for Realistic 3D Human Motion Prediction (CVPR 2025 · similarity 0.91)
- PhysDiff: Physics-Guided Human Motion Diffusion Model (ICCV 2023 · similarity 0.90)
- Existence Is Chaos: Enhancing 3D Human Motion Prediction with Uncertainty Consideration (AAAI 2024 · similarity 0.87)
- Human Motion Diffusion as a Generative Prior (ICLR 2024 · similarity 0.87)
- MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generation (AAAI 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
