Research map / Probabilistic and scientific modeling / Molecules and proteins
Trajectory Unified Transformer for Pedestrian Trajectory Prediction
Liushuai Shi, Le Wang, Sanping Zhou, Gang Hua
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 |
|---|---|---|---|
| unify social interaction and multimodal prediction in transformer without post-processing | TUTR · ETH-UCY (avg) | ADE/FDE: 0.23/0.38 -> 0.21/0.35 | Improved |
| TUTR versus methods with post-processing on ETH-UCY FDE | TUTR · ETH-UCY (avg) | FDE: 0.32 -> 0.36 | Hurt |
| TUTR versus SocialVAE+FPC on SDD FDE | TUTR · SDD | FDE: 11.72 -> 12.69 | Hurt |
| predict probabilities and trajectories in parallel (dual prediction) | TUTR · ETH-UCY (avg) | brier-ADE/FDE: 1.12/1.27 -> 0.95/1.10 | Improved |
| remove post-processing clustering | TUTR · Inference time, 80 pedestrians | seconds: 2.0939 -> 0.0533 | Improved |
| add global prediction (general motion modes) instead of learnable latent vectors (ablation) | TUTR · SDD | ADE/FDE: 17.26/34.64 -> 7.76/12.69 | Improved |
| add global prediction instead of latent vectors (ablation) | TUTR · ETH-UCY (avg) | ADE/FDE: 0.34/0.64 -> 0.21/0.36 | Improved |
| add mode-level transformer encoder (ablation) | TUTR · SDD | ADE/FDE: 8.14/13.46 -> 7.85/12.91 | Improved |
| add social-level transformer decoder (ablation) | TUTR · SDD | ADE/FDE: 7.85/12.91 -> 7.76/12.69 | Improved |
| vary number of general motion modes L (ablation) | TUTR · SDD | ADE/FDE | Mixed |
Closest papers by meaning
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.89)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (AAAI 2024 · similarity 0.87)
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.87)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.86)
- SocialMOIF: Multi-Order Intention Fusion for Pedestrian Trajectory Prediction (CVPR 2025 · similarity 0.86)
- A Unified Environmental Network for Pedestrian Trajectory Prediction (AAAI 2024 · similarity 0.86)
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.
