Research map / Detection and segmentation / 3D detection and driving
One Million Scenes for Autonomous Driving: ONCE Dataset
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Pretrain SECOND with DeepCluster on ONCE rather than nuScenes or Waymo | SECOND · KITTI downstream | moderate mAP: 66.1 -> 67.2; 66.5 -> 67.2 | Helped |
| Use voxel representation rather than pillars | 3D object detection on ONCE · ONCE test | mAP: 45.47 -> 51.90 | Helped |
| Combine point and voxel representations | PV-RCNN · ONCE test | mAP: 51.90 -> 53.85 | Helped |
| Replace anchor assignments with center assignments | SECOND · ONCE test, pedestrian | AP: 26.09% -> 51.80% | Helped |
| Replace anchor assignments with center assignments | SECOND · ONCE test, vehicle | AP: 69.71% -> 66.35% | Hurt |
| Add image segmentation scores to the point-cloud input | CenterPoints / PointPainting · ONCE test | mAP: 61.24% -> 59.78% | Hurt |
| Increase self-supervised unlabeled data from 100k to 1 million scenes | PointContrast · ONCE test | mAP: 50.75% -> 52.99% | Helped |
| Use clustering-based pretraining instead of contrastive pretraining | Self-supervised SECOND · ONCE test, U_large | mAP: 52.10% -> 54.28%; 52.99% -> 54.27% | Helped |
| Use Mean Teacher rather than self-supervised SwAV | SECOND · ONCE test, U_large | mAP: 54.28% -> 59.99% | Helped |
| Add strong augmentations to pseudo-label training | Pseudo Label · ONCE test, U_large | mAP: 49.29% -> 51.90% | Helped |
| Filter low-quality pseudo boxes using the IoU prediction head | 3DIoUMatch · ONCE test, U_large | mAP: 56.61% -> 57.43% | Helped |
| Add size and center consistency constraints | Mean Teacher (SESS) · ONCE test, U_large | mAP: 59.99% -> 58.78% | Hurt |
Closest papers by meaning
- Scalability in Perception for Autonomous Driving: Waymo Open Dataset (CVPR 2020 · similarity 0.93)
- Unsupervised Adaptation from Repeated Traversals for Autonomous Driving (NeurIPS 2022 · similarity 0.91)
- Image-to-Lidar Self-Supervised Distillation for Autonomous Driving Data (CVPR 2022 · similarity 0.91)
- DrivingStereo: A Large-Scale Dataset for Stereo Matching in Autonomous Driving Scenarios (CVPR 2019 · similarity 0.91)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.89)
- The Road Ahead in Autonomous Driving: The KITScenes Multimodal Dataset (NeurIPS 2026 · similarity 0.89)
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.