Research map / Detection and segmentation / 3D detection and driving
ForeSight: Multi-View Streaming Joint Object Detection and Trajectory Forecasting
Sandro Papais, Letian Wang, Brian Cheong, Steven L. Waslander
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 detection-based forecasting instead of tracking-based forecasting (ablation) | ForeSight · nuScenes val | minADE: 2.83 -> 1.05 | Improved |
| add historical detection query propagation (4 frames, direct) (ablation) | single-frame PETR detector · nuScenes val | mAP: 0.361 -> 0.440 | Improved |
| propagate forecast queries back to detection (bidirectional) (ablation) | ForeSight · nuScenes val | mAP: 0.440 -> 0.463 | Improved |
| add HD map to forecasting (ablation) | ForeSight · nuScenes val | minADE: 0.735 -> 0.709 | Improved |
| add HD map to forecasting (ablation) | ForeSight · nuScenes val | mAP: 0.463 -> 0.466 | Improved |
| increase history frames from 4 to 8 (ablation) | ForeSight · nuScenes val | mAP: 0.466 -> 0.467 | No clear effect |
| replace historical queries with spatial queries at equal query count (ablation) | ForeSight · nuScenes val | mAP: 0.466 -> 0.361 | Hurt |
| use 3 forecast decoder layers instead of constant-position propagation (ablation) | ForeSight · nuScenes val | mAP: 0.445 -> 0.466 | Improved |
| use 3 forecast decoder layers instead of 1 (ablation) | ForeSight · nuScenes val | minADE: 0.979 -> 0.709 | Improved |
| ForeSight vs StreamPETR on truck class | ForeSight · nuScenes val class-wise AP (V2-99) | AP: 0.628 -> 0.607 (truck); 0.366 -> 0.421 (trailer) | Mixed |
Closest papers by meaning
- Forecasting From LiDAR via Future Object Detection (CVPR 2022 · similarity 0.89)
- DINO-Foresight: Looking into the Future with DINO (NeurIPS 2025 · similarity 0.88)
- Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · similarity 0.87)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · similarity 0.86)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · similarity 0.86)
- Real-time Stereo-based 3D Object Detection for Streaming Perception (NeurIPS 2024 · similarity 0.85)
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.
