Research map / Probabilistic and scientific modeling / Molecules and proteins
Towards Metric-Agnostic Trajectory Forecasting
Markus Knoche, Daan de Geus, Bastain Leibe
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 |
|---|---|---|---|
| apply window policy to Step-NLL predictions (instead of naive output) | DONUT Step-NLL (Laplace) · Waymo val | Soft mAP: 0.3595 -> 0.5082 | Improved |
| apply distance policy to Step-NLL predictions | DONUT Step-NLL (Laplace) · Waymo val | minFDE: 1.2236 -> 1.0200 | Improved |
| apply distance policy to WTA-trained predictions | DONUT WTA · Waymo val | minFDE: 1.0520 -> 1.0824 | Hurt |
| replace WTA loss with Step-NLL when policies are applied (ablation) | DONUT · Waymo val | Soft mAP: 0.4764 -> 0.5082 | Improved |
| replace WTA loss with Traj-NLL when policies are applied (ablation) | DONUT · Waymo val | minFDE: 1.0824 -> 1.0374 | Improved |
| replace WTA loss with Step-NLL under naive evaluation (ablation) | DONUT · Waymo val | minFDE: 1.0520 -> 1.2236 | Hurt |
| replace Laplace with generalized Gaussian position distribution (with policies) (ablation) | DONUT Step-NLL · Waymo val | Soft mAP: 0.5082 -> 0.5101 | Improved |
| replace Laplace with generalized Gaussian (distance policy) (ablation) | DONUT Step-NLL · Waymo val | minFDE: 1.0200 -> 1.0103 | Improved |
| replace Laplace with Gaussian scale mixture under naive evaluation (ablation) | DONUT Step-NLL · Waymo val | Soft mAP: 0.3595 -> 0.3837 | Improved |
| replace Laplace with Gaussian scale mixture (window policy) (ablation) | DONUT Step-NLL · Waymo val | Soft mAP: 0.5082 -> 0.5053 | Hurt |
| apply policies to MTR with WTA loss (vs original NMS outputs) | MTR · Waymo val | Soft mAP: 0.4288 -> 0.4837 | Improved |
| apply distance policy to QCNet WTA predictions | QCNet · Waymo val | minFDE: 1.2254 -> 2.2805 | Hurt |
Closest papers by meaning
- Rethinking Training and Inference for Trajectory Forecasting: Linking Winner-Take-All back to GMMs (ECCV 2026 · similarity 0.89)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · similarity 0.87)
- Evaluation of Trajectory Distribution Predictions with Energy Score (ICML 2024 · similarity 0.87)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.86)
- Motion Forecasting in Continuous Driving (NeurIPS 2024 · similarity 0.86)
- Certified Human Trajectory Prediction (CVPR 2025 · 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.
