Research map / Point clouds / Segmentation and detection
PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PointDistiller knowledge distillation to a 4× compressed PointPillars student | PointPillars · KITTI val BEV moderate | mAP: 66.4 -> 69.2 | Helped |
| Add PointDistiller knowledge distillation to a 4× compressed PointPillars student | PointPillars · KITTI val 3D moderate | mAP: 58.7 -> 62.1 | Helped |
| Add PointDistiller knowledge distillation to a 4× compressed SECOND student | SECOND · KITTI val BEV moderate | mAP: 71.0 -> 73.2 | Helped |
| Add PointDistiller knowledge distillation to a 4× compressed SECOND student | SECOND · KITTI val 3D moderate | mAP: 64.7 -> 66.2 | Helped |
| Add PointDistiller knowledge distillation to a compressed PointRCNN student | PointRCNN · KITTI val BEV moderate | mAP: 74.1 -> 75.1; 73.4 -> 74.0 | Mixed |
| Add PointDistiller knowledge distillation to PointPillars | PointPillars · nuScenes 2× and 4× | mAP: 36.0 -> 36.7; 32.2 -> 32.8 | Helped |
| Add local distillation without reweighted learning (ablation) | 4× PointPillars · KITTI val BEV moderate | mAP: 66.4 -> 68.7 | Helped |
| Add reweighted learning without local distillation (ablation) | 4× PointPillars · KITTI val BEV moderate | mAP: 66.4 -> 68.4 | Helped |
| Combine local distillation with reweighted learning (ablation) | 4× PointPillars · KITTI val BEV moderate | mAP: 68.7 -> 69.2; 68.4 -> 69.2 | Helped |
| Add local distillation without reweighted learning (ablation) | 4× PointPillars · KITTI val 3D moderate | mAP: 58.7 -> 61.2 | Helped |
| Add reweighted learning without local distillation (ablation) | 4× PointPillars · KITTI val 3D moderate | mAP: 58.7 -> 60.6 | Helped |
| Combine local distillation with reweighted learning (ablation) | 4× PointPillars · KITTI val 3D moderate | mAP: 61.2 -> 62.1; 60.6 -> 62.1 | Helped |
Closest papers by meaning
- Towards Efficient 3D Object Detection with Knowledge Distillation (NeurIPS 2022 · similarity 0.91)
- Weak-to-Strong 3D Object Detection with X-Ray Distillation (CVPR 2024 · similarity 0.88)
- Voxel or Pillar: Exploring Efficient Point Cloud Representation for 3D Object Detection (AAAI 2024 · similarity 0.87)
- PointCFormer: A Relation-Based Progressive Feature Extraction Network for Point Cloud Completion (AAAI 2025 · similarity 0.87)
- Point Cloud Dataset Distillation (ICML 2025 · similarity 0.87)
- DFD: Distilling the Feature Disparity Differently for Detectors (ICML 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.