Research map / Label-efficient and robust learning / Recognition and long-tail
Forecasting Human Trajectory from Scene History
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 group trajectory bank (GTB) on top of cross-modal interaction module (ablation) | SHENet · PAV average | CS-FDE (pixels): 55.13 -> 45.13 | Helped |
| add cross-modal interaction module (TE+SE+CMT) to GTB (ablation) | SHENet · PAV average | CS-FDE (pixels): 53.07 -> 45.13 | Helped |
| add scene encoder to trajectory encoder (ablation) | CMI without GTB · PAV average | CS-ADE (pixels): 23.03 -> 21.81 | Helped |
| add cross-modal transformer over concatenated TE+SE features (ablation) | CMI without GTB · PAV average | CS-ADE (pixels): 21.81 -> 20.63 | Helped |
| GTB alone versus CMI alone (ablation) | SHENet components · PAV average | CS-ADE (pixels): 21.52 vs 20.63 | Hurt |
| set bank update threshold theta to 75% of training error rather than 25% (ablation) | SHENet · PAV average | CS-FDE (pixels): 58.28 -> 45.13 | Helped |
| set theta to 75% rather than 100% (update bank less often) (ablation) | SHENet · PAV average | CS-FDE (pixels): 54.52 -> 45.13 | Helped |
| vary initial K-medoids clusters K from 24 to 36 (ablation) | SHENet · PAV average | CS-FDE (pixels): 45.48 / 45.59 / 45.13 / 45.62 | No clear effect |
| SHENet versus YNet | SHENet · ETH/UCY average | FDE (m): 0.39 -> 0.36 | Helped |
| SHENet versus YNet on PAV | SHENet · PAV average | CS-FDE (pixels): 50.43 -> 45.13 | Helped |
| SHENet versus YNet on SDD | SHENet · SDD | FDE: 11.85 vs 13.24 | Hurt |
| remove video/scene data on ETH/UCY | SHENet · ETH/UCY average | FDE (m): 0.39 vs YNet 0.27 | Mixed |
Closest papers by meaning
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.88)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.86)
- GigaTraj: Predicting Long-term Trajectories of Hundreds of Pedestrians in Gigapixel Complex Scenes (CVPR 2024 · similarity 0.86)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.85)
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.85)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · 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.