Research map / Probabilistic and scientific modeling / Molecules and proteins
PriorMotion: Generative Class-Agnostic Motion Prediction with Raster-Vector Motion Field Priors
Kangan Qian, Jinyu Miao, Xinyu Jiao, Ziang Luo, Zheng Fu, Yining Shi, Yunlong Wang, Kun Jiang, Diange Yang
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 the vector prior encoder (VpE) to the simple generative model (ablation) | Simple.G on MotionNet · nuScenes (fast >5 m/s) | mean speed error (m): 0.9733 -> 0.9580 | Improved |
| add the vector prior encoder (VpE) to the simple generative model (ablation) | Simple.G on MotionNet · nuScenes (static cells) | mean speed error (m): 0.0255 -> 0.0318 | Hurt |
| add the raster prior encoder (RpE) to the simple generative model (ablation) | Simple.G on MotionNet · nuScenes (slow <=5 m/s) | mean speed error (m): 0.2477 -> 0.2356 | Improved |
| add the raster prior encoder (RpE) to the simple generative model (ablation) | Simple.G on MotionNet · nuScenes (fast >5 m/s) | mean speed error (m): 0.9733 -> 0.9351 | Improved |
| combine vector and raster priors (RVpE) rather than either alone (ablation) | Simple.G on MotionNet · nuScenes (fast >5 m/s) | mean speed error (m): 0.9733 -> 0.9028 | Improved |
| add dynamic-static prior modeling (DSpM) to the simple generative model (ablation) | Simple.G on MotionNet · nuScenes (static cells) | mean speed error (m): 0.0255 -> 0.0235 | Improved |
| add latent future motion generation (SGRU) on top of DSpM (ablation) | DSpG on MotionNet · nuScenes (fast >5 m/s) | mean speed error (m): 0.9564 -> 0.8943 | Improved |
| combine RVpE with DSpG (full PriorMotion) instead of either alone (ablation) | PriorMotion on MotionNet · nuScenes (fast >5 m/s) | mean speed error (m): 0.9028 -> 0.8324 | Improved |
| add PriorMotion generative prior framework to STPN backbone | MotionNet (STPN) · nuScenes (fast >5 m/s) | mean speed error (m): 0.9878 -> 0.8324 | Improved |
| add PriorMotion generative prior framework to STI backbone | BE-STI (STI) · nuScenes (fast >5 m/s) | mean speed error (m): 0.9078 -> 0.7739 | Improved |
| add PriorMotion priors on STPN for generalization to masked category | MotionNet (STPN) · nuScenes generalization (masked category) | generalization index (%): 81.6 -> 86.4 | Improved |
| add PriorMotion priors for long-distance prediction and motion stability | MotionNet (STPN) · nuScenes long distance >20 m | motion stability: 0.0267 -> 0.0088 | Improved |
Closest papers by meaning
- SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Data (ECCV 2026 · similarity 0.89)
- Self-Supervised Class-Agnostic Motion Prediction with Spatial and Temporal Consistency Regularizations (CVPR 2024 · similarity 0.88)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · similarity 0.86)
- ICP-Flow: LiDAR Scene Flow Estimation with ICP (CVPR 2024 · similarity 0.86)
- VoteFlow: Enforcing Local Rigidity in Self-Supervised Scene Flow (CVPR 2025 · similarity 0.86)
- MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction (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.
