Research map / Detection and segmentation / 3D detection and driving
GUPNet++: Geometry Uncertainty Propagation Network for Monocular 3D Object Detection
Yan Lu, Xinzhu Ma, Lei Yang, Tianzhu Zhang, Yating Liu, Qi Chu, Tong He, Yonghui Li, Wanli Ouyang
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 RoI feature maps (coordinate and category maps) to the baseline (ablation) | two-stage baseline · KITTI val Car 3D@0.7 | AP Moderate: 11.00 -> 12.44 | Improved |
| add geometry projection depth (GeP) with directly learned uncertainty (ablation) | baseline + RoI-FM · KITTI val Car 3D@0.7 | AP Moderate: 12.44 -> 12.79 | Improved |
| add vanilla uncertainty confidence UnC (no GeP) (ablation) | baseline + RoI-FM · KITTI val Car 3D@0.7 | AP Moderate: 12.44 -> 13.53 | Improved |
| add GeP when depth uncertainty is learned directly (with UnC) (ablation) | baseline + UnC · KITTI val Car 3D@0.7 | AP Easy: 19.69 -> 18.23 | Hurt |
| add geometry-guided uncertainty GeU (ablation) | GeP + UnC · KITTI val Car 3D@0.7 | AP Moderate: 13.57 -> 15.70 | Improved |
| add Hierarchical Task Learning (HTL) to GUPNet (ablation) | GeP + UnC + GeU · KITTI val Car 3D@0.7 | AP Easy: 20.86 -> 22.76 | Improved |
| replace GeU with GeU++ (uncertainty from both 2D and 3D heights) (ablation) | GUPNet · KITTI val Car 3D@0.7 | AP Moderate: 16.46 -> 16.71 | Improved |
| add HTL on top of GeU++ (ablation) | GUPNet++ (GeU++, UnC) · KITTI val Car 3D@0.7 | AP Moderate: 16.53 -> 16.71 | No clear effect |
| replace vanilla UnC with IoU-guided IoUnC (ablation) | GeU++ model · KITTI val Car 3D@0.7 | AP Moderate: 16.53 -> 17.43 | Improved |
| replace NLL with beta-NLL Laplacian loss (ablation) | GeU++ + IoUnC · KITTI val Car 3D@0.7 | AP Moderate: 17.43 -> 17.85 | Improved |
| replace 2D NMS with 3D-level NMS (ablation) | GUPNet++ · KITTI val Car 3D@0.7 | AP Moderate: 17.85 -> 19.06 | Improved |
| add deformable convolution (DCN) in DLA backbone (ablation) | GUPNet++ + NMS3D · KITTI val Car 3D@0.7 | AP Moderate: 19.06 -> 20.45 | Improved |
Later papers that compared with this method
2 later accepted papers report GUPNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Object-Scene-Camera Decomposition and Recomposition for Data Efficient Monocular 3D Object Detection (IJCV 2026 · better on 2 of 2 shared results)
- MonoWAD: Weather-Adaptive Diffusion Model for Robust Monocular 3D Object Detection (ECCV 2024 · better on 1 of 1 shared result)
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.
- MonoDETR — MonoDETR: Depth-guided Transformer for Monocular 3D Object Detection (ICCV 2023) · better on 5 of 6 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
- Geometry Uncertainty Projection Network for Monocular 3D Object Detection (ICCV 2021 · similarity 0.94)
- MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors (CVPR 2025 · similarity 0.89)
- Rethinking Geometric Depth in Monocular 3D Object Detection: A Projection-Consistent Reformulation (NeurIPS 2026 · similarity 0.89)
- MonoJSG: Joint Semantic and Geometric Cost Volume for Monocular 3D Object Detection (CVPR 2022 · similarity 0.88)
- Categorical Depth Distribution Network for Monocular 3D Object Detection (CVPR 2021 · similarity 0.87)
- MonoRUn: Monocular 3D Object Detection by Reconstruction and Uncertainty Propagation (CVPR 2021 · similarity 0.87)
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.
