Research map / Detection and segmentation / 3D detection and driving
GLENet: Boosting 3D Object Detectors with Generative Label Uncertainty Estimation
Yifan Zhang, Qijian Zhang, Zhiyu Zhu, Junhui Hou, Yixuan Yuan
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 GLENet label uncertainty into KL loss (ablation) | Voxel R-CNN with KL loss · KITTI val | 3D AP R40 moderate (%): 85.25 -> 85.76 | Improved |
| train with KL loss only (ablation) | Voxel R-CNN · KITTI val | 3D AP R40 moderate (%): 85.29 -> 85.25 | No clear effect |
| add variance voting (ablation) | Voxel R-CNN with KL loss and LU · KITTI val | 3D AP R40 moderate (%): 85.76 -> 85.91 | Improved |
| add UAQE module (ablation) | GLENet-VR · KITTI val | 3D AP R40 moderate (%): 85.91 -> 86.10 | Improved |
| use GLENet uncertainty instead of points-number heuristic (ablation) | GLENet-VR · KITTI val | 3D AP R40 moderate (%): 85.58 -> 86.10 | Improved |
| use GLENet uncertainty instead of convex hull heuristic (ablation) | GLENet-VR · KITTI val | 3D AP R40 moderate (%): 85.45 -> 86.10 | Improved |
| add occlusion augmentation to GLENet (ablation) | GLENet · KITTI val | 3D AP R40 moderate (%): 85.52 -> 86.10 | Improved |
| add context encoder to GLENet (ablation) | GLENet · KITTI val | 3D AP R40 moderate (%): 85.31 -> 86.10 | Improved |
| add absolute coordinates to GLENet input (ablation) | GLENet · KITTI val | 3D AP R40 moderate (%): 86.10 -> 85.66 | No clear effect |
| increase latent dimension from 2 to 8 (ablation) | GLENet · KITTI val | 3D AP R40 moderate (%): 84.69 -> 86.15 | Improved |
| increase sampling times from 4 to 30 (ablation) | GLENet · KITTI val | 3D AP R40 moderate (%): 85.11 -> 86.10 | Improved |
| integrate GLENet into SECOND | SECOND · KITTI val | 3D AP R40 moderate (%): 81.99 -> 84.11 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Voxel-RCNN — Voxel R-CNN: Towards High Performance Voxel-based 3D Object Detection (AAAI 2021) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Harnessing Uncertainty-aware Bounding Boxes for Unsupervised 3D Object Detection (ICCV 2025 · similarity 0.88)
- Adaptive Dual Uncertainty Optimization: Boosting Monocular 3D Object Detection under Test-Time Shifts (ICCV 2025 · similarity 0.87)
- Uncertain Object Representation for Image-Based 3D Object Perception (TPAMI 2025 · similarity 0.87)
- Not Every Side Is Equal: Localization Uncertainty Estimation for Semi-Supervised 3D Object Detection (ICCV 2023 · similarity 0.86)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.86)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.85)
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.
