Research map / 3D vision / 3D generation
Towards Immersive Human-X Interaction: A Real-Time Framework for Physically Plausible Motion Synthesis
Kaiyang Ji, Ye Shi, Zichen Jin, Kangyi Chen, Lan Xu, Yuexin Ma, Jingyi Yu, Jingya Wang
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 interaction loss (ablation) | Human-X · Inter-X | FID: 1.457 -> 0.975 | Improved |
| add foot loss (ablation) | Human-X · Inter-X | Skating: 0.239 -> 0.092 | Improved |
| add prefix loss (ablation) | Human-X · Inter-X | FID: 1.023 -> 0.975 | Improved |
| increase denoising timesteps from 10 to 100 (ablation) | Human-X · Inter-X | FID and latency: FID 0.988 -> 0.908, latency 16.9 -> 170 ms | Mixed |
| reduce denoising timesteps to 2 (ablation) | Human-X · Inter-X | FID: 0.975 -> 1.809 | Hurt |
| use history/horizon h=20, k=20 instead of default (ablation) | Human-X · Inter-X | FID: 0.975 -> 0.957 | Mixed |
| use canonical interaction representation instead of non-canonical (ablation) | Human-X · Inter-X | FID_cd: 5.737 -> 1.694 | Improved |
| add binary interaction field (ablation) | Human-X · Inter-X | FID: 1.108 -> 0.975 | Improved |
| add 6D joint rotation to the representation (ablation) | Human-X · Inter-X | FID: 0.975 -> 1.363 | Hurt |
| use 8 decoder layers instead of 2 (ablation) | Human-X · Inter-X | FID: 2.729 -> 0.975 | Improved |
| use DistilBERT instead of CLIP text encoder (ablation) | Human-X · Inter-X | FID: 1.023 -> 0.975 | Improved |
| add interaction field of 1x1 (pelvis only) instead of six joints (ablation) | Human-X · Inter-X | FID: 1.095 -> 0.975 | Hurt |
Closest papers by meaning
- Habitat 3.0: A Co-Habitat for Humans, Avatars, and Robots (ICLR 2024 · similarity 0.89)
- InterControl: Zero-shot Human Interaction Generation by Controlling Every Joint (NeurIPS 2024 · similarity 0.88)
- Ready-to-React: Online Reaction Policy for Two-Character Interaction Generation (ICLR 2025 · similarity 0.88)
- InterMimic: Towards Universal Whole-Body Control for Physics-Based Human-Object Interactions (CVPR 2025 · similarity 0.88)
- ReGenNet: Towards Human Action-Reaction Synthesis (CVPR 2024 · similarity 0.88)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · 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.
