Research map / Point clouds / Segmentation and detection
Exploring Active 3D Object Detection from a Generalization Perspective
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 |
|---|---|---|---|
| select samples using label conciseness, feature representativeness, and geometric balance together | CRB active 3D detection · KITTI val, Average Moderate | 3D AP: 63.23 -> 67.81 | Helped |
| add label conciseness sampling (ablation) | CRB active 3D detection · KITTI val, 1,000 boxes | 3D AP, Moderate: $58.27_{\pm1.04}$ -> $64.56_{\pm1.39}$ | Helped |
| add representative prototype selection (ablation) | CRB active 3D detection · KITTI val, 1,000 boxes | 3D AP, Moderate: $58.27_{\pm1.04}$ -> $61.65_{\pm1.95}$ | Helped |
| add geometric balance sampling alone (ablation) | CRB active 3D detection · KITTI val, 1,000 boxes | 3D AP, Moderate: $58.27_{\pm1.04}$ -> $56.12_{\pm12.74}$ | Hurt |
| add geometric balance to Cls and Rps (ablation) | CRB active 3D detection · KITTI val, 1,000 boxes | 3D AP, Moderate: $62.81_{\pm1.31}$ -> $65.86_{\pm1.21}$ | Helped |
| add representative prototype selection to Cls and Gpdb (ablation) | CRB active 3D detection · KITTI val, 1,000 boxes | 3D AP, Moderate: $64.70_{\pm1.07}$ -> $65.86_{\pm1.21}$ | Helped |
| use K-means rather than GMM or K-means++ for prototype selection (ablation) | CRB Rps module · KITTI val, Moderate | 3D mAP: ~1.5% | No clear effect |
| vary KDE bandwidth among 5, 7, and 9 rather than use the preferred setting (ablation) | CRB Gpdb module · KITTI val | 3D mAP: 2% | No clear effect |
| use KDE bandwidth 3 (ablation) | CRB Gpdb module · KITTI val | 3D mAP | Mixed |
| vary the candidate thresholds K1 and K2 (ablation) | CRB active 3D detection · KITTI val, Moderate and Hard | Average mAP: 3.28%; 2.81% | No clear effect |
| use CRB selection with SECOND instead of the compared generic active learning approaches | SECOND 3D detector · KITTI val, 3% queried boxes | 3D mAP: 4.7% | Helped |
Closest papers by meaning
- STONE: A Submodular Optimization Framework for Active 3D Object Detection (NeurIPS 2024 · similarity 0.93)
- AS-Det: Active Sampling for Adaptive 3D Object Detection in Point Clouds (AAAI 2025 · similarity 0.91)
- Uncertainty Meets Diversity: A Comprehensive Active Learning Framework for Indoor 3D Object Detection (CVPR 2025 · similarity 0.90)
- iDet3D: Towards Efficient Interactive Object Detection for LiDAR Point Clouds (AAAI 2024 · similarity 0.90)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.89)
- MixSup: Mixed-grained Supervision for Label-efficient LiDAR-based 3D Object Detection (ICLR 2024 · 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.