Research map / Video / Video generation
SOSControl: Enhancing Human Motion Generation Through Saliency-Aware Symbolic Orientation and Timing Control
Ho Yin Au, Junkun Jiang, Jie Chen
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 |
|---|---|---|---|
| add SMS-based training data processing (ablation) | SOSControl · HumanML3D test | L2-Rot6D / FID: 0.499 -> 0.325 L2-Rot6D; 13.494 -> 3.892 FID | Improved |
| add SMS-based training data processing (ablation) | SOSControl · HumanML3D test | SOS-Acc: 0.991 -> 0.988 | No clear effect |
| add SMS-based training data processing (ablation) | OmniControl · HumanML3D test | SOS-Acc: 0.225 -> 0.873 | Improved |
| add test-time iterative optimization (ablation) | SOSControl · HumanML3D test | SOS-Acc: 0.531 -> 0.988 | Improved |
| add diffusion-time iterative optimization (ablation) | SOSControl · HumanML3D test | SOS-Acc: 0.531 -> 0.535 | No clear effect |
| add diffusion-time on top of test-time optimization (ablation) | SOSControl · HumanML3D test | FID / MMD: FID 4.209 -> 3.892; MMD 6.125 -> 6.199 | Mixed |
| add diffusion-time iterative optimization (ablation) | GMD 1-stage · HumanML3D test | SOS-Acc: 0.111 -> 0.113 | No clear effect |
| use text-only MDM instead of SOS-conditioned model | MDM · HumanML3D test | FID: FID 2.592 vs 3.892; SOS-Acc 0.151 vs 0.988 | Mixed |
Closest papers by meaning
- MotionCharacter: Fine-Grained Motion Controllable Human Video Generation (AAAI 2026 · similarity 0.86)
- KinMo: Kinematic-aware Human Motion Understanding and Generation (ICCV 2025 · similarity 0.86)
- MotionCtrl: A Real-time Controllable Vision-Language-Motion Model (ICCV 2025 · similarity 0.86)
- Enhanced Fine-Grained Motion Diffusion for Text-Driven Human Motion Synthesis (AAAI 2024 · similarity 0.85)
- MMM: Generative Masked Motion Model (CVPR 2024 · similarity 0.85)
- ONE-SHOT: Compositional Human-Environment Video Synthesis via Spatial-Decoupled Motion Injection and Hybrid Context Integration (NeurIPS 2026 · similarity 0.85)
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.
