Research map / Detection and segmentation / 3D detection and driving
Canonical Voting: Towards Robust Oriented Bounding Box Detection in 3D Scenes
Check 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 |
|---|---|---|---|
| Use LCC canonical voting and back-projection checking instead of direct offset voting (ablation) | Canonical Voting · ScanNet and SceneNN, separate models | mAP50: 0.0 -> 21.7; 0.0 -> 26.8 | Helped |
| Add LCC back-projection checking (ablation) | Canonical Voting · ScanNet, separate models | mAP50: 8.1 -> 21.7 | Helped |
| Add LCC back-projection checking (ablation) | Canonical Voting · SceneNN, separate models | mAP50: 10.0 -> 26.8 | Helped |
| Add objectness-weighted voting rather than assigning every point weight one (ablation) | Canonical Voting · ScanNet, separate models | mAP50: 12.2 -> 21.7 | Helped |
| Add objectness-weighted voting rather than assigning every point weight one (ablation) | Canonical Voting · SceneNN, separate models | mAP50: 21.4 -> 26.8 | Helped |
| Train separately for each category rather than jointly | Canonical Voting · ScanNet | mAP50: 15.4 -> 21.7 | Helped |
| Train separately for each category rather than jointly | Canonical Voting · SceneNN | mAP50: 21.1 -> 26.8 | Helped |
| Generate orientations through canonical voting rather than direct orientation regression | Canonical Voting · ScanNet | orientation error: 10.8 vs 76.2 | Helped |
| Increase orientation sampling resolution K (ablation) | Canonical Voting · ScanNet | mAP50 | No clear effect |
| Use canonical voting for detecting occluded or partial objects | Canonical Voting · ScanNet, severely occluded objects | AR50 | Helped |
Closest papers by meaning
- 3D Bounding Box Estimation Using Deep Learning and Geometry (CVPR 2017 · similarity 0.88)
- ImVoteNet: Boosting 3D Object Detection in Point Clouds With Image Votes (CVPR 2020 · similarity 0.86)
- Prototypical VoteNet for Few-Shot 3D Point Cloud Object Detection (NeurIPS 2022 · similarity 0.86)
- Back-Tracing Representative Points for Voting-Based 3D Object Detection in Point Clouds (CVPR 2021 · similarity 0.86)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (CVPR 2024 · similarity 0.85)
- OpenBox: Annotate Any Bounding Boxes in 3D (NeurIPS 2025 · similarity 0.84)
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.