Research map / Probabilistic and scientific modeling / Molecules and proteins
On Exposing the Challenging Long Tail in Future Prediction of Traffic Actors
Osama Makansi, Özgün Çiçek, Yassine Marrakchi, Thomas Brox
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 difficulty-based contrastive loss (ablation) | Traj++ EWTA · ETH-UCY (Top 1%) | min-FDE: 0.87 -> 0.71 | Improved |
| add difficulty-based contrastive loss (ablation) | Traj++ EWTA · nuScenes bird's-eye (Top 1%) | min-FDE: 1.02 -> 0.85 | Improved |
| add difficulty-based contrastive loss (ablation) | FLN-RPN · nuScenes egocentric (Top 1%) | min-FDE: 36.16 -> 27.49 | Improved |
| add difficulty-based contrastive loss (zero-shot transfer from nuScenes) (ablation) | FLN-RPN · Waymo (Top 1%) | min-FDE: 27.32 -> 24.09 | Improved |
| add difficulty-based contrastive loss (ablation) | Traj++ EWTA · ETH-UCY (All) | min-FDE: 0.32 -> 0.32 | No clear effect |
| add difficulty-based contrastive loss (ablation) | FLN-RPN · Waymo (All) | min-FDE: 6.39 -> 6.49 | No clear effect |
| replace cVAE with EWTA multi-hypothesis loss (ablation) | Trajectron++ · ETH-UCY (All) | min-FDE: 0.41 -> 0.32 | Improved |
| add resampling of challenging scenarios | Traj++ EWTA · ETH-UCY (All) | min-FDE: 0.32 -> 0.53 | Hurt |
| add loss reweighting | Traj++ EWTA · ETH-UCY (All) | min-FDE: 0.32 -> 0.56 | Hurt |
| add resampling of challenging scenarios | FLN-RPN · Waymo (All) | min-FDE: 6.39 -> 10.48 | Hurt |
| add resampling of challenging scenarios | FLN-RPN · nuScenes egocentric (Top 1%) | min-FDE: 36.16 -> 21.62 | Improved |
| add LDAM loss | FLN-RPN · nuScenes egocentric (Top 1%) | min-FDE: 36.16 -> 31.13 | Improved |
Closest papers by meaning
- Generating Useful Accident-Prone Driving Scenarios via a Learned Traffic Prior (CVPR 2022 · similarity 0.87)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.85)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.85)
- Overcoming Challenges of Long-Horizon Prediction in Driving World Models (NeurIPS 2025 · similarity 0.85)
- Generating Traffic Scenarios via In-Context Learning to Learn Better Motion Planner (AAAI 2025 · similarity 0.85)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · 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.
