Research map / Detection and segmentation / 3D detection and driving
Disentangling Monocular 3D Object Detection
Andrea Simonelli, Samuel Rota Bulo, Lorenzo Porzi, Manuel Lopez-Antequera, Peter Kontschieder
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 signed IoU loss to RetinaNet (ablation) | RetinaNet · KITTI3D Car 2D, Moderate | AP|R40: 83.74 -> 84.05 | Improved |
| add disentanglement to signed IoU loss (ablation) | RetinaNet + IoU · KITTI3D Car 2D, Moderate | AP|R40: 84.05 -> 85.38 | Improved |
| replace direct 10D regression with entangled 3D box loss (ablation) | Regression · KITTI3D Car 3D, Moderate | AP|R40: 0.90 -> 2.70 | Improved |
| add signed IoU 2D loss and 3D confidence (ablation) | 3D BB · KITTI3D Car 3D, Moderate | AP|R40: 2.70 -> 6.10 | Improved |
| add signed IoU 2D loss and 3D confidence (ablation) | Regression · KITTI3D Car 3D, Moderate | AP|R40: 0.90 -> 1.70 | Improved |
| add disentangling to the 3D box loss (ablation) | 3D BB · KITTI3D Car 3D, Moderate | AP|R40: 2.70 -> 3.00 | Improved |
| combine all contributions (MonoDIS vs 3D BB w/ IoUDIS, 3DConf) (ablation) | 3D BB w/ IoUDIS, 3DConf · KITTI3D Car 3D, Moderate | AP|R40: 6.10 -> 7.60 | Improved |
| rank detections by 3D confidence instead of 2D confidence (ablation) | MonoDIS · KITTI3D val Car 3D, Moderate | AP|R11: 4.91 -> 14.98 | Improved |
| rank detections by 3D confidence instead of random confidence (ablation) | MonoDIS · KITTI3D val Car 3D, Moderate | AP|R11: 2.37 -> 14.98 | Improved |
| use entangled 3D box loss vs regression, effect on 2D detection (ablation) | Regression · KITTI3D Car 2D, Moderate | AP|R40: 73.20 -> 77.10 | Improved |
| train on larger training split | MonoDIS · KITTI3D test Car 3D, Moderate | AP|R40: 4.89 -> 7.94 | Improved |
| use disentangled loss instead of entangled in toy optimization (ablation) | 3D box loss · toy KITTI3D box fitting | convergence | 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.
- MonoGRNet — MonoGRNet: A Geometric Reasoning Network for Monocular 3D Object Localization (AAAI 2019) · better on 3 of 3 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
- Disentangling Monocular 3D Object Detection: From Single to Multi-Class Recognition (TPAMI 2020 · similarity 0.98)
- Homography Loss for Monocular 3D Object Detection (CVPR 2022 · similarity 0.91)
- Monocular 3D Object Detection with Bounding Box Denoising in 3D by Perceiver (ICCV 2023 · similarity 0.90)
- Rethinking Geometric Depth in Monocular 3D Object Detection: A Projection-Consistent Reformulation (NeurIPS 2026 · similarity 0.90)
- MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors (CVPR 2025 · similarity 0.89)
- MonoDiff: Monocular 3D Object Detection and Pose Estimation with Diffusion Models (CVPR 2024 · similarity 0.88)
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.
