Research map / Detection and segmentation / 3D detection and driving
Forecasting From LiDAR via Future Object Detection
Neehar Peri, Jonathon Luiten, Mengtian Li, Aljoša Ošep, Laura Leal-Taixé, Deva Ramanan
Check 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 |
|---|---|---|---|
| predict future detections and backcast them to current detections instead of forecasting forward from current detections | CenterPoint-based end-to-end car forecasting · nuScenes val cars, K=1 | mAP_f: 31.5 -> 33.5 | Improved |
| predict multiple future hypotheses and evaluate the top 5 | FutureDet · nuScenes val cars | mAP_f: 33.5 -> 35.6 | Improved |
| use a PointPillars backbone instead of the default backbone (ablation) | FutureDet · nuScenes val cars, K=1 | mAP_f: 33.5 -> 32.8 | Hurt |
| use a PointPillars backbone instead of the default backbone (ablation) | FutureDet · nuScenes val cars, K=5 | mAP_f: 35.6 -> 36.1 | Improved |
| add road masks as an input channel (ablation) | FutureDet · nuScenes val cars, K=1 | mAP_f: 33.5 -> 33.1 | Hurt |
| add road masks as an input channel (ablation) | FutureDet · nuScenes val pedestrians, K=1 | mAP_f: 26.9 -> 26.5 | No clear effect |
| use a PointPillars backbone instead of the default backbone (ablation) | FutureDet · nuScenes val pedestrians, K=1 | mAP_f: 26.9 -> 21.3 | Hurt |
| use a constant position forecast instead of a motion-aware forecast | constant-position baseline · nuScenes val cars | ADE@60: 0.38 (Constant Position); 0.74 (Fast and Furious*) | Mixed |
| use a constant position forecast instead of forecasting movement | constant-position baseline · nuScenes val cars | mAP_f: 22.1 -> 31.5 | Hurt |
| use trajectory-level copy-paste augmentation to oversample linear and nonlinear trajectories | FutureDet training · nuScenes val cars | AP_f | Improved |
Closest papers by meaning
- DeTra: A Unified Model for Object Detection and Trajectory Forecasting (ECCV 2024 · similarity 0.90)
- Argoverse 2: Next Generation Datasets for Self-Driving Perception and Forecasting (NeurIPS 2021 · similarity 0.89)
- UnO: Unsupervised Occupancy Fields for Perception and Forecasting (CVPR 2024 · similarity 0.89)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.87)
- Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · similarity 0.86)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 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.
