Research map / Label-efficient and robust learning / Knowledge distillation
Representation Disparity-aware Distillation for 3D Object Detection
Yanjing Li, Sheng Xu, Mingbao Lin, Jihao Yin, Baochang Zhang, Xianbin Cao
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 |
|---|---|---|---|
| replace Hint (full neck feature distillation) with RDD region selection (ablation) | CP-Voxel-S · nuScenes | mAP: 52.9 -> 57.1 | Improved |
| use Hint region selection instead of no distillation (ablation) | CP-Voxel-S · nuScenes | mAP: 54.0 -> 52.9 | Hurt |
| add feature-level RDD loss alone (ablation) | CP-Voxel-S · nuScenes | mAP: 54.0 -> 56.8 | Improved |
| add logit-level RDD loss alone (ablation) | CP-Voxel-S · nuScenes | mAP: 54.0 -> 57.0 | Improved |
| combine feature-level and logit-level RDD losses (ablation) | CP-Voxel-S · nuScenes | mAP: 57.0 -> 57.1 | Improved |
| set lambda to 0 so all region pairs are distilled equally (ablation) | RDD · nuScenes CP-Voxel-S | mAP | Hurt |
| enlarge distillation loss weights alpha1 and alpha2 beyond 0.2 (ablation) | RDD · nuScenes CP-Voxel-S | mAP | Hurt |
| distill with RDD versus previous 3D KD (PP Logit KD) | CP-Voxel-XXS · nuScenes | mAP: 47.9 -> 49.4 | Improved |
| distill with RDD versus PP Logit KD on the smallest nuScenes voxel student | CP-Voxel-XS · nuScenes | mAP: 53.5 -> 54.0 | Improved |
| distill CP-Voxel-S with RDD, student surpasses its teacher | CP-Voxel-S · nuScenes | mAP: 56.6 -> 57.1 | Improved |
| apply 2D feature distillation baselines (GID, LAD) to 3D detectors | CP-Voxel-S · nuScenes | mAP: 54.0 -> 54.1 (GID-L), 53.0 (LAD) | No clear effect |
| distill with RDD versus PointDistiller on KITTI | PointPillars-S · KITTI val moderate | mAP@R40: 62.3 -> 63.0 | Improved |
Closest papers by meaning
- Towards Efficient 3D Object Detection with Knowledge Distillation (NeurIPS 2022 · similarity 0.93)
- Advancing 3D Object Detection With Depth-Aware Spatial Knowledge Distillation (TPAMI 2025 · similarity 0.92)
- DFD: Distilling the Feature Disparity Differently for Detectors (ICML 2024 · similarity 0.92)
- Beyond the Limitation of Monocular 3D Detector via Knowledge Distillation (ICCV 2023 · similarity 0.91)
- PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection (CVPR 2023 · similarity 0.91)
- DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection (AAAI 2025 · 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.
