Research map / Video / Video generation
Motion Synthesis with Sparse and Flexible Keyjoint Control
Inwoo Hwang, Jinseok Bae, Donggeun Lim, Young Min Kim
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 |
|---|---|---|---|
| decompose into keyjoint trajectory stage then full-body completion stage (ablation) | goal-driven motion synthesis · Goal-driven all scenarios | distance to goal (m): 0.206 -> 0.093 | Improved |
| decompose into two stages for implicit hand-to-head control (ablation) | DNO-style noise optimization · Hand-to-head | constraint error: 0.0093 -> 0.0036 | Improved |
| use time-agnostic arc-length alignment loss instead of explicit uniform timestamps (ablation) | trajectory control · Time-agnostic control | constraint error: 0.4827 -> 0.0256 | Improved |
| decompose into keyjoint stage, reducing sampling cost (ablation) | motion control runtime · Runtime | time (s): 41.2 -> 7.1 | Improved |
| use DDIM sampling with 10 or 5 steps instead of DDPM (ablation) | keyjoint pipeline · HumanML3D sparse control r=30 | FID: 0.224 -> 0.254 (DDIM10), 0.338 (DDIM5) | Mixed |
| make control signals sparser with larger frame interval (ablation) | keyjoint pipeline · HumanML3D | FID: 0.127 (r=1) -> 0.263 (r=60) | Hurt |
| perturb sparse control signals with out-of-distribution noise up to 5 cm (ablation) | keyjoint pipeline · HumanML3D r=30 | control error (m): 0.038 -> 0.043 | No clear effect |
| use keyjoint decomposition vs prior baselines under sparse control | motion synthesis · HumanML3D sparse control | control error (m): 0.111 -> 0.036 (vs OmniControl) | Improved |
Closest papers by meaning
- Programmable Motion Generation for Open-Set Motion Control Tasks (CVPR 2024 · similarity 0.89)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.89)
- Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance (NeurIPS 2025 · similarity 0.87)
- Motion Prompting: Controlling Video Generation with Motion Trajectories (CVPR 2025 · similarity 0.86)
- Follow-Your-Click: Open-domain Regional Image Animation via Motion Prompts (AAAI 2025 · similarity 0.86)
- DisPose: Disentangling Pose Guidance for Controllable Human Image Animation (ICLR 2025 · similarity 0.86)
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.
