Research map / Probabilistic and scientific modeling / Molecules and proteins
Den-TP: A Density-Balanced Data Curation and Evaluation Framework for Trajectory Prediction
Ruining Yang, Yi Xu, Yun Fu, Lili Su
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 |
|---|---|---|---|
| Use Den-TP selection instead of random selection at 50% | HiVT-64 on 50% subset · Argoverse 1 val | minADE: 0.750 -> 0.706 | Improved |
| Add submodular selection on top of density partitioning (ablation) | Den-TP · Argoverse 1 val (HiVT-64) | minADE: 0.729 -> 0.706 | Improved |
| Add density partitioning on top of submodular selection (ablation) | Den-TP · Argoverse 1 val (HiVT-64) | minADE: 0.724 -> 0.706 | Improved |
| Density-balanced Den-TP selection vs retaining all high-density scenarios plus random fill | HiVT-64 on 50% subset · Argoverse 1 val | minADE: 0.724 -> 0.706 | Improved |
| Reweight high-density samples (full data) | HiVT-64 · Argoverse 1 val | minADE: 0.715 vs 0.695 full data | No clear effect |
| Epoch-wise resampling instead of random fixed subset | HiVT-64 at 95k · Argoverse 1 val | minADE: 0.750 -> 0.752 | Hurt |
| Train on 50% Den-TP subset instead of full data | HiVT-64 · Argoverse 1 val | minADE: 0.695 -> 0.706 | No clear effect |
| Train on 50% Den-TP subset instead of full data | QCNet · Argoverse 2 val | minADE: 0.724 -> 0.754 | Hurt |
| Use Den-TP subset selected by HiVT-64 to train other models vs random subset (50%) | DeMo · Argoverse 2 val | minADE: 0.756 -> 0.704 | Improved |
| Use Den-TP subset vs Herding at 50% | HiVT-64/HPNet · Argoverse 1 val | minADE: 0.728 -> 0.706 (HiVT-64) | Improved |
| Train on Den-TP subset instead of full data on high-density scenes | HiVT/HPNet · Argoverse 1 high-density subset | JADE: 1.741 -> 1.677 | Improved |
| Train on Den-TP subset instead of full data on full validation | HiVT/HPNet · Argoverse 1 full val | JADE: 0.816 -> 0.819 | No clear effect |
Closest papers by meaning
- C2F-TP: A Coarse-to-Fine Denoising Framework for Uncertainty-Aware Trajectory Prediction (AAAI 2025 · similarity 0.88)
- Certified Human Trajectory Prediction (CVPR 2025 · similarity 0.87)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.87)
- BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving (AAAI 2024 · similarity 0.86)
- Evaluation of Trajectory Distribution Predictions with Energy Score (ICML 2024 · similarity 0.86)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 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.
