Research map / Point clouds / Segmentation and detection
MoDAR: Using Motion Forecasting for 3D Object Detection in Point Cloud Sequences
Yingwei Li, Charles R. Qi, Yin Zhou, Chenxi Liu, Dragomir Anguelov
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 |
|---|---|---|---|
| Add MoDAR from past and future frames | CenterPoint · Waymo Open Dataset val | L2 3D mAPH: 58.1 -> 69.2 (+11.1) | Improved |
| Add MoDAR from past and future frames | SWFormer · Waymo Open Dataset val | L2 3D mAPH: 66.2 -> 74.7 (+8.5) | Improved |
| Add MoDAR from past and future frames | 3-frame SWFormer · Waymo Open Dataset val | L2 3D mAPH: 69.9 -> 76.4 (+6.5) | Improved |
| Use MP++ instead of stationary or constant-velocity prediction (ablation) | CenterPoint + MoDAR · Waymo Open Dataset val, vehicle L2 | APH: 73.0 -> 73.5 | Improved |
| Use six MP++ trajectories instead of one (ablation) | CenterPoint + MoDAR · Waymo Open Dataset val, vehicle L2 | APH: 73.5 -> 73.2 | No clear effect |
| Use a stationary predictor for MoDAR (ablation) | CenterPoint + MoDAR · Waymo Open Dataset val, vehicle L2, stationary vehicles | APH: 60.5 -> 72.0 | Improved |
| Use a stationary predictor rather than LiDAR-only detection (ablation) | CenterPoint + MoDAR · Waymo Open Dataset val, vehicle L2, very fast vehicles | APH: 79.9 -> 77.6 | Hurt |
| Add MoDAR location to LiDAR-only detection (ablation) | CenterPoint · Waymo Open Dataset val, vehicle L2 | APH: 64.2 -> 69.7 (+5.5) | Improved |
| Add MoDAR location to LiDAR-only detection (ablation) | CenterPoint · Waymo Open Dataset val, pedestrian L2 | APH: 51.9 -> 61.2 (+10.7) | Improved |
| Add size, heading, class, and scores to location-only MoDAR (ablation) | CenterPoint + MoDAR · Waymo Open Dataset val, vehicle L2 | APH: 69.7 -> 73.2 | Improved |
| Add MoDAR to LiDAR detection (ablation) | 3-frame SWFormer · Waymo Open Dataset val, vehicle L2, 50m+ | APH: 45.6 -> 58.5 (+12.9) | Improved |
| Use input-level early fusion rather than weighted-box late fusion (ablation) | 3-frame SWFormer with LiDAR and MoDAR · Waymo Open Dataset val, vehicle L2 | APH: 70.4 -> 77.0 | Improved |
Closest papers by meaning
- MGTANet: Encoding Sequential LiDAR Points Using Long Short-Term Motion-Guided Temporal Attention for 3D Object Detection (AAAI 2023 · similarity 0.90)
- PTT: Point-Trajectory Transformer for Efficient Temporal 3D Object Detection (CVPR 2024 · similarity 0.88)
- Learning Scene Dynamics from Point Cloud Sequences (IJCV 2022 · similarity 0.88)
- Streaming Object Detection for 3-D Point Clouds (ECCV 2020 · similarity 0.88)
- TREND: Unsupervised 3D Representation Learning via Temporal Forecasting for LiDAR Perception (NeurIPS 2025 · similarity 0.88)
- LiDAR-Net: A Real-scanned 3D Point Cloud Dataset for Indoor Scenes (CVPR 2024 · similarity 0.87)
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.
