Research map / Detection and segmentation / 3D detection and driving
WeakM3D: Towards Weakly Supervised Monocular 3D Object Detection
Liang Peng, Senbo Yan, Boxi Wu, Zheng Yang, Xiaofei He, Deng Cai
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 geometric alignment loss (ablation) | WeakM3D with only center loss · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 0.11/0.08 -> 49.09/37.79 | Improved |
| Add learning disentanglement (ablation) | WeakM3D with only center loss · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 0.11/0.08 -> 16.16/11.55 | Improved |
| Add ray tracing to geometric alignment (ablation) | WeakM3D with learning disentanglement and geometric alignment · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 51.36/45.04 -> 55.31/48.35 | Improved |
| Balance point-wise losses by point density (ablation) | WeakM3D with learning disentanglement, geometric alignment, and ray tracing · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 55.31/48.35 -> 58.20/50.16 | Improved |
| Learn dimensions rather than freeze them (ablation) | WeakM3D · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 58.20/50.16 -> 56.24/48.37 | Hurt |
| Use 2D masks rather than 2D boxes to select LiDAR points (ablation) | WeakM3D · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 55.94/47.52 -> 58.20/50.16 | Improved |
| Use ResNet50 instead of ResNet18 (ablation) | WeakM3D · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 50.92/44.78 -> 58.20/50.16 | Improved |
| Apply weak supervision rather than full supervision on the basic WeakM3D baseline | WeakM3D baseline · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 53.61/44.78 -> 58.20/50.16 | Improved |
| Apply weak supervision rather than full supervision with PatchNet | PatchNet · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 64.31/57.94 -> 59.41/54.10 | Hurt |
| Use WeakM3D instead of M3D-RPN | Waymo LEVEL 1 · Waymo, AP_BE V, 30-50m | AP_BE V: 3.92 -> 7.73 | Improved |
| Use WeakM3D instead of Autolabels | KITTI car weakly supervised detection · KITTI val, car, IoU=0.5, R40, Easy | AP_BE V/AP_3D: 50.51/38.31 -> 58.20/50.16 | Improved |
Closest papers by meaning
- VSRD: Instance-Aware Volumetric Silhouette Rendering for Weakly Supervised 3D Object Detection (CVPR 2024 · similarity 0.91)
- Weakly Supervised Monocular 3D Detection with a Single-View Image (CVPR 2024 · similarity 0.91)
- WeakPCSOD: Overcoming the Bias of Box Annotations for Weakly Supervised Point Cloud Salient Object Detection (AAAI 2024 · similarity 0.90)
- Weak-to-Strong 3D Object Detection with X-Ray Distillation (CVPR 2024 · similarity 0.89)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.89)
- LabelAny3D: Label Any Object 3D in the Wild (NeurIPS 2025 · similarity 0.88)
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.
