Research map / Image generation / Fast diffusion sampling
BCDiff: Bidirectional Consistent Diffusion for Instantaneous Trajectory Prediction
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 unidirectional guidance from predicted unobserved trajectories to future generation (ablation) | BCDiff (Trajectron++) · ETH/UCY | ADE/FDE (m): 0.36/0.70 -> 0.33/0.60 | Helped |
| add unidirectional guidance from predicted unobserved trajectories to future generation (ablation) | BCDiff (Trajectron++) · SDD | ADE/FDE (m): 13.07/22.88 -> 12.57/21.26 | Helped |
| make guidance bidirectional (future guides unobserved too) (ablation) | BCDiff (Trajectron++) · ETH/UCY | ADE/FDE (m): 0.33/0.60 -> 0.32/0.57 | Helped |
| make guidance bidirectional (future guides unobserved too) (ablation) | BCDiff (Trajectron++) · SDD | ADE/FDE (m): 12.57/21.26 -> 11.96/19.88 | Helped |
| add gating between predicted and observed guidance (ablation) | BCDiff (Trajectron++) · ETH/UCY | ADE/FDE (m): 0.32/0.57 -> 0.29/0.53 | Helped |
| add gating between predicted and observed guidance (ablation) | BCDiff (Trajectron++) · SDD | ADE/FDE (m): 11.96/19.88 -> 11.56/19.32 | Helped |
| increase predicted unobserved points up to 4 (ablation) | BCDiff (Trajectron++) · ETH/UCY | ADE/FDE (m): 0.31/0.55 -> 0.29/0.53 | Helped |
| increase predicted unobserved points beyond 4 to 6 (ablation) | BCDiff (Trajectron++) · ETH/UCY | ADE/FDE (m): 0.29/0.53 -> 0.34/0.60 | Hurt |
| replace direct decoder with BCDiff decoder under 2-frame observation | BCDiff · ETH/UCY AVG (4 backbones) | ADE/FDE (m): 0.36/0.70 -> 0.29/0.53 (Trajectron++) | Helped |
| BCDiff vs MOE on ZARA2 with Trajectron++ | BCDiff (Trajectron++) · ETH/UCY ZARA2 | ADE/FDE (m): 0.17/0.32 -> 0.18/0.33 | No clear effect |
Closest papers by meaning
- Pedestrian Trajectory Prediction with Missing Data: Datasets, Imputation, and Benchmarking (NeurIPS 2024 · similarity 0.90)
- Towards Predicting Any Human Trajectory In Context (NeurIPS 2025 · similarity 0.88)
- ProDiff: Prototype-Guided Diffusion for Minimal Information Trajectory Imputation (ICML 2025 · similarity 0.88)
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.88)
- MS-TIP: Imputation Aware Pedestrian Trajectory Prediction (ICML 2024 · similarity 0.86)
- SocialCVAE: Predicting Pedestrian Trajectory via Interaction Conditioned Latents (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.