Research map / Detection and segmentation / Tracking
Towards Trajectory Forecasting From Detection
Zhang Pu, Lei Bai, Yuning Wang, Jianwu Fang, Jianru Xue, Nanning Zheng, Wanli Ouyang
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 |
|---|---|---|---|
| train on tracked detections instead of ground-truth trajectories (ablation) | Distance-association baseline · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 148.8 -> 118.5 | Improved |
| add learned short-term affinity in place of distance association (ablation) | Model 1 · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 118.5 -> 117.5 | Improved |
| add long-term motion cues to the affinity network (ablation) | Model 2 · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 117.5 -> 115.0 | Improved |
| add affinity-aware state update (ASU) (ablation) | Model 3 · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 281.7 -> 272.4 | Improved |
| add multiple state aggregation (MSA) (ablation) | Model 3 · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 281.7 -> 273.2 | Improved |
| add both ASU and MSA (ablation) | Model 3 · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 281.7 -> 269.3 | Improved |
| use affinity score as uncertainty input to ASU (ablation) | Model 3 · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 281.7 -> 279.5 | Improved |
| use multi-dimensional affinity feature instead of scalar score (ablation) | ASU · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 279.5 -> 276.0 | Improved |
| modulate affinity features with GRU_aff instead of per-step feature (ablation) | ASU · nuScenes vehicle nonlinear (CenterPoint, rec@0.6) | nl_fde@3s: 276.0 -> 272.4 | Improved |
| add ASU+MSA on MATF social interaction model (ablation) | MATF · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 108.9 -> 105.9 | Improved |
| add ASU+MSA on SR-LSTM social interaction model (ablation) | SR-LSTM · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 108.0 -> 104.9 | Improved |
| add ASU+MSA on STGAT social interaction model (ablation) | STGAT · nuScenes vehicle (CenterPoint, rec@0.6) | fde@3s: 114.4 -> 110.6 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PnPNet — PnPNet: End-to-End Perception and Prediction With Tracking in the Loop (CVPR 2020) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Whose Track Is It Anyway? Improving Robustness to Tracking Errors With Affinity-Based Trajectory Prediction (CVPR 2022 · similarity 0.90)
- Certified Human Trajectory Prediction (CVPR 2025 · similarity 0.85)
- DeMo: Decoupling Motion Forecasting into Directional Intentions and Dynamic States (NeurIPS 2024 · similarity 0.85)
- OOSTraj: Out-of-Sight Trajectory Prediction With Vision-Positioning Denoising (CVPR 2024 · similarity 0.85)
- LaKD: Length-agnostic Knowledge Distillation for Trajectory Prediction with Any Length Observations (NeurIPS 2024 · similarity 0.84)
- HPNet: Dynamic Trajectory Forecasting with Historical Prediction Attention (CVPR 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.
