Research map / Probabilistic and scientific modeling / Molecules and proteins
Three Steps to Multimodal Trajectory Prediction: Modality Clustering, Classification and Synthesis
Jianhua Sun, Yuxuan Li, Hao-Shu Fang, Cewu Lu
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 |
|---|---|---|---|
| cluster deep trajectory features instead of raw coordinates (ablation) | PCCSNet · ETH/UCY average | ADE: 0.24 -> 0.21 | Improved |
| cluster deep trajectory features instead of raw coordinates (ablation) | PCCSNet · ETH/UCY average | FDE: 0.45 -> 0.42 | Improved |
| use HAC instead of K-means (ablation) | PCCSNet · ETH/UCY average | time per run (min): 0.7 -> 18 | No clear effect |
| use GMM instead of K-means (ablation) | PCCSNet · ETH/UCY average | time per run (min): 0.7 -> 90 | No clear effect |
| vary K in K-means (100, 200, 500, 1000) (ablation) | PCCSNet · ETH/UCY average | ADE: 0.22 (K=100), 0.21 (K=200), 0.21 (K=500), 0.22 (K=1000) | Mixed |
| shift clustering weights toward historical representation (2:1, 3:1 vs 1:1) (ablation) | PCCSNet · ETH/UCY average | FDE: 0.42 -> 0.47 -> 0.49 | Hurt |
| shift clustering weights toward future representation (1:2, 1:3 vs 1:1) (ablation) | PCCSNet · ETH/UCY average | FDE: 0.42 -> 0.42 -> 0.43 | No clear effect |
| add Modality Loss (ablation) | PCCSNet · ETH/UCY average | ADE: 0.221 -> 0.208 | Improved |
| add Modality Loss (ablation) | PCCSNet · ETH/UCY average | FDE: 0.443 -> 0.422 | Improved |
| increase Modality Loss radius from 1 to 2 (ablation) | PCCSNet · ETH/UCY average | ADE: 0.208 -> 0.215 | Hurt |
| relax Modality Loss speed threshold from 10% to 20% (ablation) | PCCSNet · ETH/UCY average | ADE: 0.208 -> 0.216 | Hurt |
| add modality synthesis (ablation) | PCCSNet · ETH/UCY average | ADE: 0.23 -> 0.21 | Improved |
Closest papers by meaning
- Modality Exploration, Retrieval and Adaptation for Trajectory Prediction (TPAMI 2023 · similarity 0.90)
- ModeSeq: Taming Sparse Multimodal Motion Prediction with Sequential Mode Modeling (CVPR 2025 · similarity 0.86)
- Can Language Beat Numerical Regression? Language-Based Multimodal Trajectory Prediction (CVPR 2024 · similarity 0.85)
- Multi-modal Knowledge Distillation-based Human Trajectory Forecasting (CVPR 2025 · similarity 0.84)
- GeoMM: On Geodesic Perspective for Multi-modal Learning (CVPR 2025 · similarity 0.84)
- MotionMap: Representing Multimodality in Human Pose Forecasting (CVPR 2025 · similarity 0.84)
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.
