Research map / Detection and segmentation / 3D detection and driving
Perspective-Invariant 3D Object Detection
Ao Liang, Lingdong Kong, Dongyue Lu, Youquan Liu, Jian Fang, Huaici Zhao, Wei Tsang Ooi
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 Pi3DET-Net adaptation to PV-RCNN trained on nuScenes (vehicle to quadruped) | PV-RCNN · Pi3DET Quadruped | AP 3D @0.7: 33.55 -> 46.36 | Improved |
| add Pi3DET-Net adaptation versus target-only training | PV-RCNN · nuScenes -> Pi3DET Quadruped | AP 3D @0.7: 40.24 -> 46.36 | Improved |
| add Random Platform Jitter (ablation) | Voxel RCNN · Pi3DET Vehicle -> Drone | AP BEV @0.7: 52.85 -> 60.20 | Improved |
| add Virtual Platform Pose (ablation) | Voxel RCNN · Pi3DET Vehicle -> Drone | AP BEV @0.7: 52.85 -> 59.83 | Improved |
| combine RPJ and VPP (ablation) | Voxel RCNN · Pi3DET Vehicle -> Drone | AP BEV @0.7: 52.85 -> 64.52 | Improved |
| add KL probabilistic feature alignment (ablation) | Voxel RCNN · Pi3DET Vehicle -> Drone | AP 3D @0.7: 41.50 -> 46.83 | Improved |
| add Geometry-Aware Transformation Descriptor (ablation) | Voxel RCNN · Pi3DET Vehicle -> Drone | AP 3D @0.7: 46.83 -> 47.75 | Improved |
| add Geometry-Aware Transformation Descriptor (ablation) | Voxel RCNN · Pi3DET Vehicle -> Quadruped | AP BEV @0.7: 55.72 -> 55.54 | No clear effect |
| add Random Platform Jitter to top vehicle detectors | Part A2 · Pi3DET Vehicle | AP 3D @0.7: 48.23 -> 47.56 | Hurt |
| add RPJ to PV-RCNN++ | PV-RCNN++ · Pi3DET Quadruped | AP 3D @0.7: 22.35 -> 31.09 | Improved |
| adapt from nuScenes to KITTI with Pi3DET-Net without RPJ | SECOND-IoU · nuScenes -> KITTI | average AP 3D: 43.43 -> 46.45 | 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.
- ST3D++ — ST3D: Self-Training for Unsupervised Domain Adaptation on 3D Object Detection (CVPR 2021) · better on 1 of 1 shared result
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
- PerspectiveNet: 3D Object Detection from a Single RGB Image via Perspective Points (NeurIPS 2019 · similarity 0.88)
- ObjectNet3D: A Large Scale Database for 3D Object Recognition (ECCV 2016 · similarity 0.87)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.86)
- Universal Object Detection with Large Vision Model (IJCV 2023 · similarity 0.85)
- MonoDGP: Monocular 3D Object Detection with Decoupled-Query and Geometry-Error Priors (CVPR 2025 · similarity 0.84)
- Omni3D: A Large Benchmark and Model for 3D Object Detection in the Wild (CVPR 2023 · similarity 0.84)
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.
