Research map / Detection and segmentation / 3D detection and driving
Approaching Outside: Scaling Unsupervised 3D Object Detection from 2D Scene
Ruiyang Zhang, Hu Zhang, Hang Yu, Zhedong Zheng
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 image-derived 3D boxes to LiDAR pseudo-labels (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 11.8 -> 16.2 | Improved |
| use only image boxes beyond 10m instead of all image boxes (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 13.8 -> 16.2 | Improved |
| raise the image-box distance threshold from 10m to 15m (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 16.2 -> 14.3 | Hurt |
| add adaptive sampling strategy (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 16.2 -> 18.4 | Improved |
| add weak model aggregation (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 16.2 -> 18.5 | Improved |
| combine adaptive sampling and weak model aggregation (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 18.4 -> 19.8 | Improved |
| use combined volume and distance sampling instead of volume only (ablation) | adaptive sampling · nuScenes (0-80m) | AP_BEV: 17.1 -> 18.4 | Improved |
| start weak model aggregation later (round 8 vs round 3) (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 16.3 -> 18.5 | Improved |
| decrease aggregation coefficient lambda from 0.999 to 0.9 (ablation) | LiSe · nuScenes (0-80m) | AP_BEV: 18.5 -> 16.0 | Hurt |
| use image-only pseudo boxes versus LiDAR-only at far range (ablation) | LiSe · nuScenes (50-80m) | AP_BEV: 1.5 -> 0.6 | Hurt |
| apply LiSe instead of MODEST on Lyft | LiSe · Lyft (50-80m) | AP_BEV: 27.0 -> 46.6 | Improved |
Closest papers by meaning
- Towards Unsupervised Object Detection From LiDAR Point Clouds (CVPR 2023 · similarity 0.92)
- Unsupervised Object Detection With LIDAR Clues (CVPR 2021 · similarity 0.92)
- OpenSight: A Simple Open-Vocabulary Framework for LiDAR-Based Object Detection (ECCV 2024 · similarity 0.91)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.90)
- GBlobs: Explicit Local Structure via Gaussian Blobs for Improved Cross-Domain LiDAR-based 3D Object Detection (CVPR 2025 · similarity 0.89)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · 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.
