Research map / Detection and segmentation / 3D detection and driving
Reinforced Axial Refinement Network for Monocular 3D Object Detection
Lijie Liu, Chufan Wu, Jiwen Lu, Lingxi Xie, Jie Zhou, Qi Tian
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 |
|---|---|---|---|
| use discrete refinement output instead of continuous regression (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 1.81 -> 18.25 | Improved |
| use parameter-aware image enhancement instead of simple fusion (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 0.40 -> 18.25 | Improved |
| use parameter-aware module instead of direct projection (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 10.88 -> 18.25 | Improved |
| use axial coordinate instead of world coordinate (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 5.34 -> 18.25 | Improved |
| output a single refinement operation (and use RL) instead of all at once (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 2.27 -> 18.25 | Improved |
| optimize with reinforcement learning (ablation) | RAR-Net (MonoGRNet) · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 13.96 -> 18.25 | Improved |
| add RAR-Net refinement | Deep3DBox · KITTI val 2, Easy, IoU=0.7 | 3D AP (%): 5.85 -> 14.25 | Improved |
| add RAR-Net refinement | MonoGRNet · KITTI val 1, Easy, IoU=0.7 | 3D AP (%): 13.84 -> 18.25 | Improved |
| add RAR-Net refinement | M3D-RPN · KITTI test, Easy | 3D AP (%): 14.76 -> 16.37 | Improved |
| add RAR-Net refinement | D4LCN (uses depth data) · KITTI | 3D AP (%): +0.51% AP | Improved |
| add RAR-Net refinement | Deep3DBox · KITTI val 2, Easy | AOS (%): 98.59 -> 98.61 | No clear effect |
| add RAR-Net refinement | Deep3DBox · KITTI bird's eye view val 2, Easy, IoU=0.7 | AP (%): 9.99 -> 14.38 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- M3D-RPN — M3D-RPN: Monocular 3D Region Proposal Network for Object Detection (ICCV 2019) · better on 6 of 6 shared results
- GS3D — GS3D: An Efficient 3D Object Detection Framework for Autonomous Driving (CVPR 2019) · better on 4 of 4 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
- Triangulation Learning Network: From Monocular to Stereo 3D Object Detection (CVPR 2019 · similarity 0.88)
- RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection (NeurIPS 2026 · similarity 0.87)
- MonoTTA: Fully Test-Time Adaptation for Monocular 3D Object Detection (ECCV 2024 · similarity 0.87)
- Monocular 3D Object Detection with Decoupled Structured Polygon Estimation and Height-Guided Depth Estimation (AAAI 2020 · similarity 0.86)
- MonoTAKD: Teaching Assistant Knowledge Distillation for Monocular 3D Object Detection (CVPR 2025 · similarity 0.86)
- MonoPlace3D: Learning 3D-Aware Object Placement for 3D Monocular Detection (CVPR 2025 · similarity 0.85)
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.
