Research map / Detection and segmentation / 3D detection and driving
Geometry-Based Distance Decomposition for Monocular 3D Object Detection
Xuepeng Shi, Qi Ye, Xiaozhi Chen, Chuangrong Chen, Zhixiang Chen, Tae-Kyun Kim
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 |
|---|---|---|---|
| replace direct distance regression with geometry-based decomposition (ablation) | MonoRCNN baseline · KITTI val (Car) | AP3D R40 Mod: 10.52 -> 10.97 | Improved |
| use decomposition instead of eight projected corners plus physical size (ablation) | MonoRCNN · KITTI val (Car) | AP3D R40 Mod: 8.96 -> 10.97 | Improved |
| add uncertainty-aware regression loss on H and h_rec (ablation) | MonoRCNN (D) · KITTI val (Car) | AP3D R40 Mod: 10.97 -> 12.00 | Improved |
| add uncertainty-aware regression loss on H and h_rec (ablation) | MonoRCNN (D) · KITTI val (Car) | APBEV R40 Mod: 15.52 -> 17.23 | Improved |
| sort boxes by score/(f H sigma_hrec) instead of score (ablation) | MonoRCNN (D+U) · KITTI val (Car) | AP3D R40 Mod: 12.00 -> 13.19 | Improved |
| sort boxes by score/(f H sigma_hrec) instead of score (ablation) | MonoRCNN (D+U) · KITTI val (Car) | AP3D R40 Easy: 16.94 -> 16.61 | Hurt |
| sort boxes by score/(f H sigma_hrec) instead of score (ablation) | MonoRCNN (D+U) · KITTI val (Car) | APBEV R40 Mod: 17.23 -> 19.22 | Improved |
| replace predicted physical height with groundtruth at inference | MonoRCNN · KITTI val (Car), self-consistency | AP3D R40 Mod: 12.00 -> 10.82 | Hurt |
| replace predicted physical size with groundtruth at inference | keypoint-based K · KITTI val (Car), self-consistency | AP3D R40 Mod: 8.96 -> 9.51 | Improved |
| use decomposition instead of M3D-RPN distance regression in cross-dataset test | MonoRCNN · nuScenes cross-test | mean distance error (m): 2.75 -> 2.39 | Improved |
| evaluate far objects in cross-dataset test | MonoRCNN vs M3D-RPN · nuScenes cross-test, 40m+ | mean distance error (m): 10.73 -> 8.65 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- M3D-RPN — M3D-RPN: Monocular 3D Region Proposal Network for Object Detection (ICCV 2019) · better on 5 of 5 shared results
- MonoPair — MonoPair: Monocular 3D Object Detection Using Pairwise Spatial Relationships (CVPR 2020) · 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
- MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors (CVPR 2025 · similarity 0.90)
- Distance-Normalized Unified Representation for Monocular 3D Object Detection (ECCV 2020 · similarity 0.90)
- Monocular 3D Object Detection with Decoupled Structured Polygon Estimation and Height-Guided Depth Estimation (AAAI 2020 · similarity 0.90)
- Depth-discriminative Metric Learning for Monocular 3D Object Detection (NeurIPS 2023 · similarity 0.90)
- Objects Are Different: Flexible Monocular 3D Object Detection (CVPR 2021 · similarity 0.90)
- Disentangling Monocular 3D Object Detection (ICCV 2019 · 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.
