Research map / Detection and segmentation / 3D detection and driving
LargeAD: Large-Scale Cross-Sensor Data Pretraining for Autonomous Driving
Lingdong Kong, Xiang Xu, Youquan Liu, Jun Cen, Runnan Chen, Wenwei Zhang, Liang Pan, Kai Chen, Ziwei Liu
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 VFM-based superpixels (SEEM) to camera-to-LiDAR distillation (ablation) | LargeAD · nuScenes LP | mIoU: 38.80 -> 43.00 | Improved |
| add superpoint temporal consistency (STC) to C2L (ablation) | LargeAD · nuScenes LP | mIoU: 38.80 -> 40.45 | Improved |
| add STC on top of VFM (ablation) | LargeAD · nuScenes LP | mIoU: 43.00 -> 44.01 | Improved |
| add point-to-segment (P2S) regularization on top of VFM (ablation) | LargeAD · nuScenes LP | mIoU: 43.00 -> 43.35 | Improved |
| add P2S on top of VFM and STC (ablation) | LargeAD · nuScenes LP | mIoU: 44.01 -> 44.95 | Improved |
| add cross-dataset pretraining (CDP) (ablation) | LargeAD · nuScenes LP | mIoU: 44.95 -> 46.13 | Improved |
| add cross-dataset pretraining (CDP) (ablation) | LargeAD · SemanticKITTI 1% | mIoU: 46.63 -> 49.55 | Improved |
| replace SLIC superpixels with SEEM (ablation) | LargeAD · nuScenes LP | mIoU: 40.89 -> 44.95 | Improved |
| replace SEEM with SAM superpixels (ablation) | LargeAD · nuScenes LP | mIoU: 44.95 -> 43.94 | Hurt |
| replace SEEM with SAM superpixels (ablation) | LargeAD · nuScenes 25% | mIoU: 68.41 -> 69.08 | Improved |
| use SLIC to SEEM superpixels in SLidR (ablation) | SLidR · nuScenes LP | mIoU: 38.80 -> 43.00 | Improved |
| use larger 2D distillation backbone (ViT-L instead of ResNet-50) | LargeAD · nuScenes LP | mIoU: 46.13 -> 48.71 | 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.
- Seal — SEAL: Self-supervised Embodied Active Learning using Exploration and 3D Consistency (NeurIPS 2021) · 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
- Image-to-Lidar Self-Supervised Distillation for Autonomous Driving Data (CVPR 2022 · similarity 0.90)
- Enhanced Spatiotemporal Consistency for Image-to-LiDAR Data Pretraining (TPAMI 2025 · similarity 0.90)
- Multi-Modal Data-Efficient 3D Scene Understanding for Autonomous Driving (TPAMI 2025 · similarity 0.90)
- VLM2Scene: Self-Supervised Image-Text-LiDAR Learning with Foundation Models for Autonomous Driving Scene Understanding (AAAI 2024 · similarity 0.89)
- UniPAD: A Universal Pre-training Paradigm for Autonomous Driving (CVPR 2024 · similarity 0.88)
- LiMoE: Mixture of LiDAR Representation Learners from Automotive Scenes (CVPR 2025 · 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.
