Research map / Detection and segmentation / 3D detection and driving
ODAM: Object Detection, Association, and Mapping Using Posed RGB Video
Kejie Li, Daniel DeTone, Yu Fan (Steven) Chen, Minh Vo, Ian Reid, Hamid Rezatofighi, Chris Sweeney, Julian Straub, Richard Newcombe
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 GNN data association instead of 3D IoU matching (ablation) | ODAM · ScanNet data association | matching accuracy: 0.85 -> 0.88 | Improved |
| add monocular 3D attributes as GNN input (ablation) | ODAM · ScanNet data association | matching accuracy: 0.84 -> 0.88 | Improved |
| add frame-to-model association (detection fusion) over frame-to-frame (ablation) | ODAM · ScanNet data association | matching accuracy: 0.86 -> 0.88 | Improved |
| use super-quadric instead of ellipsoid (ablation) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 20.7 -> 29.7 | Improved |
| use super-quadric instead of 3D cuboid (ablation) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 27.2 -> 29.7 | Improved |
| add multi-view optimization over averaged monocular predictions (ablation) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 23.9 -> 29.7 | Improved |
| add category-conditioned scale prior to multi-view optimization (ablation) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 22.1 -> 29.7 | Improved |
| multi-view optimization without scale prior vs no optimization (ablation) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 23.9 -> 22.1 | Hurt |
| ODAM versus Vid2CAD and MOLTR (full systems) | ODAM · ScanNet/Scan2CAD IoU>0.25 | F1: 55.2 -> 61.5 (Vid2CAD); 55.0 -> 61.5 (MOLTR) | Improved |
| ODAM versus Vid2CAD and MOLTR (full systems) | ODAM · ScanNet/Scan2CAD IoU>0.5 | F1: 16.5 -> 29.7 (Vid2CAD); 16.0 -> 29.7 (MOLTR) | Improved |
| RGB-only ODAM versus RGB-D VoteNet on thin objects | ODAM vs VoteNet · ScanNet per-class F1 | F1: 53.0 -> 12.5 (door); 40.7 -> 12.8 (window) | Hurt |
| RGB-only ODAM versus RGB-D VoteNet on isolated large objects | ODAM vs VoteNet · ScanNet per-class F1 | F1: 60.0 -> 65.6 (table); 88.1 -> 87.7 (bed) | Mixed |
Later papers that compared with this method
3 later accepted papers report ODAM as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Explaining Object Detectors via Collective Contribution of Pixels (CVPR 2026 · better on 1 of 1 shared result)
- FFAM: Feature Factorization Activation Map for Explanation of 3D Detectors (NeurIPS 2024 · better on 5 of 5 shared results)
- Pixel-Aligned Recurrent Queries for Multi-View 3D Object Detection (ICCV 2023 · better on 1 of 1 shared result)
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
- 2D-Driven 3D Object Detection in RGB-D Images (ICCV 2017 · similarity 0.88)
- Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild (CVPR 2023 · similarity 0.87)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 2024 · similarity 0.87)
- Graph R-CNN: Towards Accurate 3D Object Detection with Semantic-Decorated Local Graph (ECCV 2022 · similarity 0.87)
- Objectron: A Large Scale Dataset of Object-Centric Videos in the Wild With Pose Annotations (CVPR 2021 · similarity 0.87)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.86)
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.
