Research map / 3D vision / Human pose estimation
Scene-aware Human Motion Forecasting via Mutual Distance Prediction
Chaoyue Xing, Wei Mao, Miaomiao Liu
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 mutual distance prediction to scene-aware forecasting (ablation) | Scene-only forecasting model · HUMANISE seen scenes | Path error (mm) mean: 52.6 -> 45.6 | Improved |
| add scene input to motion-only model (ablation) | Motion only · HUMANISE seen scenes | Path error (mm) mean: 57.0 -> 52.6 | Improved |
| add per-vertex signed distance (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 46.2 -> 45.6 | Improved |
| add per-basis point distance (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 50.6 -> 45.6 | Improved |
| use SDF volume scene input instead of point cloud (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 49.3 -> 45.6 | Improved |
| use SDF volume scene input instead of occupancy volume (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 55.3 -> 45.6 | Improved |
| use per-basis point distance instead of BPS encoding (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 53.9 -> 45.6 | Improved |
| use stage-wise training instead of end-to-end from scratch (ablation) | Mutual distance model · HUMANISE seen scenes | Path error (mm) mean: 51.5 -> 45.6 | Improved |
| replace predicted contact map with predicted mutual distance (ground-truth input version) | STAG / ContAware · GTA-IM | Path error (mm) mean: 40.0 -> 22.4 | Improved |
| mutual distance vs contact map on pose error with GT input | STAG w/GT · HUMANISE seen scenes | Pose error 1.0s: 117.6 vs 122.0 | Mixed |
Closest papers by meaning
- Contact-aware Human Motion Forecasting (NeurIPS 2022 · similarity 0.94)
- Long-term Human Motion Prediction with Scene Context (ECCV 2020 · similarity 0.91)
- Harmonizing Stochasticity and Determinism: Scene-responsive Diverse Human Motion Prediction (NeurIPS 2024 · similarity 0.89)
- Expressive Forecasting of 3D Whole-Body Human Motions (AAAI 2024 · similarity 0.89)
- MotionMap: Representing Multimodality in Human Pose Forecasting (CVPR 2025 · similarity 0.88)
- Multimodal Sense-Informed Forecasting of 3D Human Motions (CVPR 2024 · similarity 0.88)
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.
