Research map / Detection and segmentation / 3D detection and driving
General Geometry-aware Weakly Supervised 3D Object Detection
Guowen Zhang, Junsong Fan, Liyi Chen, Zhaoxiang Zhang, Zhen Lei, Lei Zhang
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 point-to-box alignment loss to BPL (ablation) | GGA · KITTI val Car IoU=0.7 Easy | AP: 7.92 -> 26.39 | Improved |
| add semantic ratio loss to BPL (ablation) | GGA · KITTI val Car IoU=0.7 Easy | AP: 7.92 -> 42.21 | Improved |
| add PAL to BPL+SRL (ablation) | GGA · KITTI val Car IoU=0.7 Easy | AP: 42.21 -> 56.02 | Improved |
| add SRL to BPL+PAL (ablation) | GGA · KITTI val Car IoU=0.7 Easy | AP: 26.39 -> 56.02 | Improved |
| use SRL or PAL without BPL (ablation) | GGA · KITTI val | convergence | Hurt |
| vary semantic ratio prior value from 1.9 to 2.9 (ablation) | SRL · KITTI val Car IoU=0.7 Easy | AP: 44.05 (2.8) to 56.02 (2.4) | Mixed |
| train detectors on GGA pseudo-labels instead of WeakM3D (monocular) | MonoDETR · KITTI val Car IoU=0.7 Easy | AP3D: 17.06 -> 21.18 | Improved |
| GGA versus WeakMono3D on hard monocular samples | GGA · KITTI val Car IoU=0.5 Hard | AP BEV/AP3D: 40.07/36.34 vs 37.09/34.22 | Hurt |
| GGA 2D-box supervision versus full 3D-box supervision for LiDAR detector | PointRCNN · KITTI test BEV Pedestrian | AP BEV: 49.43 -> 12.07 | Hurt |
| GGA with 2D boxes versus 3D-box supervised DSS on SUN-RGBD | FCAF3D · SUN-RGBD val | mAP@0.25: 42.1 -> 48.5 | Improved |
Closest papers by meaning
- Weakly Supervised 3D Object Detection via Multi-Level Visual Guidance (ECCV 2024 · similarity 0.91)
- 2D-Driven 3D Object Detection in RGB-D Images (ICCV 2017 · similarity 0.90)
- Transferable Semi-Supervised 3D Object Detection From RGB-D Data (ICCV 2019 · similarity 0.90)
- 3D Bounding Box Estimation Using Deep Learning and Geometry (CVPR 2017 · similarity 0.90)
- VSRD: Instance-Aware Volumetric Silhouette Rendering for Weakly Supervised 3D Object Detection (CVPR 2024 · similarity 0.88)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 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.
