Research map / Point clouds / Point cloud classification
AD-PT: Autonomous Driving Pre-Training with Large-scale Point Cloud Dataset
Jiakang Yuan, Bo Zhang, Xiangchao Yan, Botian Shi, Tao Chen, Yikang LI, Yu Qiao
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 |
|---|---|---|---|
| Initialize PV-RCNN++ with AD-PT instead of MeanTeacher pre-training | PV-RCNN++ · Waymo validation (L2) | Overall AP / APH: 70.62 / 68.14 -> 71.55 / 69.23 | Improved |
| Initialize CenterPoint with AD-PT rather than training from scratch | CenterPoint · nuScenes validation, 100% training data | NDS: 64.5 -> 65.48 | Improved |
| Initialize PV-RCNN with AD-PT rather than ProposalContrast | PV-RCNN · KITTI validation, 100% training data | Moderate mAP (AP3D): 72.92 -> 73.01 | Improved |
| Add object-size re-scaling (ablation) | AD-PT data preparation · nuScenes | mAP: 36.26 -> 39.72 | Improved |
| Add LiDAR-beam re-sampling (ablation) | AD-PT data preparation · nuScenes | mAP: 36.26 -> 41.35 | Improved |
| Add both object-size re-scaling and LiDAR-beam re-sampling (ablation) | AD-PT data preparation · nuScenes | mAP: 36.26 -> 43.11 | Improved |
| Add unknown-aware instance learning (ablation) | AD-PT · nuScenes | mAP: 43.11 -> 43.92 | Improved |
| Add consistency loss to unknown-aware instance learning (ablation) | AD-PT · nuScenes | mAP: 43.92 -> 44.99 | Improved |
| Use 256 selected features instead of 128 (ablation) | AD-PT unknown-aware instance learning · Waymo (L2) | Overall AP: 67.71 -> 68.33 | Improved |
| Increase selected features from 256 to 512 (ablation) | AD-PT unknown-aware instance learning · Waymo (L2) | Overall AP: 68.33 -> 67.93 | Hurt |
| Use matching threshold 0.3 instead of 0.1 (ablation) | AD-PT unknown-aware instance learning · Waymo (L2) | Overall AP: 67.90 -> 68.33 | Improved |
| Increase matching threshold from 0.3 to 0.5 (ablation) | AD-PT unknown-aware instance learning · Waymo (L2) | Overall AP: 68.33 -> 67.82 | Hurt |
Closest papers by meaning
- Visual Point Cloud Forecasting enables Scalable Autonomous Driving (CVPR 2024 · similarity 0.89)
- One Million Scenes for Autonomous Driving: ONCE Dataset (NeurIPS 2021 · similarity 0.89)
- Self-Supervised Class-Agnostic Motion Prediction with Spatial and Temporal Consistency Regularizations (CVPR 2024 · similarity 0.87)
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025 · similarity 0.87)
- nuReasoning: A Reasoning-Centric Dataset and Benchmark for Long-Tail Autonomous Driving (NeurIPS 2026 · similarity 0.86)
- SubjectDrive: Scaling Generative Data in Autonomous Driving via Subject Control (AAAI 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.
