Research map / Detection and segmentation / 3D detection and driving
Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild
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 virtual depth (ablation) | Cube R-CNN · Omni3D | AP3D: 17.3 -> 23.3; +6% | Helped |
| Add scale augmentation with virtual depth (ablation) | Cube R-CNN · Omni3D | AP3D: +3.1% | Helped |
| Use scale augmentation without virtual depth (ablation) | Cube R-CNN · Omni3D | AP3D: -2.5% | Hurt |
| Add 3D uncertainty for prediction scoring and training (ablation) | Cube R-CNN · Omni3D subsets | AP3D: 17.2 -> 23.3; about +6% | Helped |
| Use uncertainty to scale training samples but not at inference (ablation) | Cube R-CNN · Omni3D | AP3D: +1.6% | Helped |
| Add the all-variable 3D loss (ablation) | Cube R-CNN · Omni3D | AP3D: 20.2 -> 23.3; +3.1% | Helped |
| Add disentangled 3D losses (ablation) | Cube R-CNN · Omni3D | AP3D: 22.6 -> 23.3; +0.7% | Helped |
| Replace RPN objectness with IoUness (ablation) | Cube R-CNN · Omni3D | AP3D: 22.2 -> 23.3; +1.1% | Helped |
| Add virtual camera | PGD · Omni3D | AP3D: 11.2 -> 15.4; +4.2% | Helped |
| Train on Omni3D IN rather than SUN RGB-D | Cube R-CNN · SUN RGB-D test | AP3D: 34.7 -> 35.4 | Helped |
| Train on the integrated Omni3D benchmark rather than a single dataset | Cube R-CNN · KITTI test | AP3D: 37.1 -> 37.8; +5.3% | Helped |
| Train on the full 98 categories rather than the 50 frequent categories | Cube R-CNN · Omni3D | AP3D: 23.3 to 14.1 | Hurt |
Closest papers by meaning
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.91)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.90)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.90)
- UNION: Unsupervised 3D Object Detection using Object Appearance-based Pseudo-Classes (NeurIPS 2024 · similarity 0.90)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.90)
- ImageNet3D: Towards General-Purpose Object-Level 3D Understanding (NeurIPS 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.