Research map / Architectures and efficiency / Attention and transformers
Real-Time Motion Prediction via Heterogeneous Polyline Transformer with Relative Pose Encoding
Zhejun Zhang, Alexander Liniger, Christos Sakaridis, Fisher Yu, Luc Van Gool
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 |
|---|---|---|---|
| replace pairwise-relative input with scene-centric representation (ablation) | HPTR · WOMD valid | soft mAP: 0.399 -> 0.246 | Hurt |
| replace pairwise-relative input with scene-centric representation (ablation) | HPTR · WOMD valid | minFDE: 1.145 -> 1.687 | Hurt |
| use agent-centric Wayformer baseline (60 epochs) instead of HPTR (ablation) | HPTR · WOMD valid | soft mAP: 0.399 -> 0.378 | Hurt |
| train the agent-centric WF baseline for 100 epochs (ablation) | WF baseline · WOMD valid | soft mAP: 0.378 -> 0.397 | Improved |
| replace lower triangular attention with full attention (early fusion) (ablation) | HPTR · WOMD valid | soft mAP: 0.399 -> 0.386 | Hurt |
| replace lower triangular attention with diagonal attention (late fusion) (ablation) | HPTR · WOMD valid | soft mAP: 0.399 -> 0.387 | Hurt |
| replace lower triangular attention with diagonal followed by full attention (hierarchical fusion) (ablation) | HPTR · WOMD valid | soft mAP: 0.399 -> 0.391 | Hurt |
| add projected RPE to projected q,k,v instead of k,v (ablation) | HPTR · WOMD valid | soft mAP: 0.401 -> 0.397 | No clear effect |
| concatenate RPE to k,v instead of adding projected RPE (ablation) | HPTR · WOMD valid | soft mAP: 0.401 -> 0.395 | No clear effect |
| concatenate RPE to q,k,v (ablation) | HPTR · WOMD valid | soft mAP: 0.401 -> 0.396 | No clear effect |
| remove q,k,v bias (ablation) | HPTR · WOMD valid | soft mAP: 0.401 -> 0.396 | No clear effect |
| use end-to-end HPTR instead of GoRela on AV2 (ablation) | HPTR vs GoRela · AV2 test | brier-minFDE6: 2.01 -> 2.03 | Hurt |
Closest papers by meaning
- SEPT: Towards Efficient Scene Representation Learning for Motion Prediction (ICLR 2024 · similarity 0.87)
- Motion Forecasting in Continuous Driving (NeurIPS 2024 · similarity 0.86)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · similarity 0.86)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 2024 · similarity 0.86)
- DualAD: Disentangling the Dynamic and Static World for End-to-End Driving (CVPR 2024 · similarity 0.86)
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (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.
