Research map / Detection and segmentation / 3D detection and driving
Transferable Semi-Supervised 3D Object Detection From RGB-D Data
Yew Siang Tang, Gim Hee Lee
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 |
|---|---|---|---|
| use relaxed reprojection loss with upper bound 1.5x instead of direct reprojection loss (ablation) | Ours + R · SUN-RGBD val CC semi-supervised | mAP: 25.6 -> 37.2 | Improved |
| enlarge the reprojection upper bound from 1.5x to 2.0x (ablation) | Ours + R · SUN-RGBD val CC semi-supervised | mAP: 37.2 -> 30.8 | Hurt |
| add the BoxPC classification objective to the BoxPC Fit network (ablation) | Ours + BoxPC · SUN-RGBD val CC semi-supervised | mAP: 32.1 -> 40.6 | Improved |
| add the BoxPC regression objective to the BoxPC Fit network (ablation) | Ours + BoxPC · SUN-RGBD val CC semi-supervised | mAP: 39.8 -> 40.6 | Improved |
| use combined box-point cloud input representation instead of independent features (ablation) | Ours + BoxPC · SUN-RGBD val CC semi-supervised | mAP: 34.4 -> 40.6 | Improved |
| remove the one-hot class vector from the baseline | FPN* · SUN-RGBD val CC semi-supervised | mAP: 24.5 -> 30.1 | Improved |
| add relaxed reprojection loss and BoxPC Fit network to the baseline (ablation) | FPN* w/o OneHot · SUN-RGBD val CC semi-supervised | mAP: 30.1 -> 41.1 | Improved |
| add box prior loss P (ablation) | Ours + BoxPC + R · SUN-RGBD val CC semi-supervised | mAP: 41.1 -> 41.8 | Improved |
| add BoxPC + R + P to the baseline on Cars | FPN · KITTI val CC semi-supervised | AP: 0.07% -> 59.95% | Improved |
| add BoxPC + R + P to the baseline on Pedestrians | FPN · KITTI val CC semi-supervised | AP: 48.03% -> 67.49% | Improved |
| add BoxPC + R + P to the baseline on Cyclists | FPN · KITTI val CC semi-supervised | AP: 41.89% -> 51.83% | Improved |
| add BoxPC Refine to fully-supervised FPN* | FPN* · SUN-RGBD val fully-supervised | mAP: 51.5 -> 53.2 | Improved |
Closest papers by meaning
- General Geometry-aware Weakly Supervised 3D Object Detection (ECCV 2024 · similarity 0.90)
- Leveraging Temporal Cues for Semi-Supervised Multi-View 3D Object Detection (CVPR 2025 · similarity 0.90)
- DetMatch: Two Teachers Are Better than One for Joint 2D and 3D Semi-Supervised Object Detection (ECCV 2022 · similarity 0.90)
- Semi-supervised 3D Object Detection with PatchTeacher and PillarMix (AAAI 2024 · similarity 0.88)
- WeakPCSOD: Overcoming the Bias of Box Annotations for Weakly Supervised Point Cloud Salient Object Detection (AAAI 2024 · similarity 0.88)
- Dual-Perspective Knowledge Enrichment for Semi-supervised 3D Object Detection (AAAI 2024 · 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.
