Research map / Detection and segmentation / 3D detection and driving
Benchmarking Robustness of 3D Object Detection to Common Corruptions
Yinpeng Dong, Caixin Kang, Jinlai Zhang, Zijian Zhu, Yikai Wang, Xiao Yang, Hang Su, Xingxing Wei, Jun Zhu
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 LiDAR-camera fusion rather than camera-only detection | 3D object detectors · nuScenes-C | mAPcor: 10.26; 10.68; 18.60; 22.79 -> 56.99; 58.73; 61.03 | Improved |
| Use LiDAR-camera fusion rather than LiDAR-only detection | 3D object detectors · nuScenes-C | mAPcor: 23.42; 40.37; 49.81 -> 56.99; 58.73; 61.03 | Improved |
| Use LiDAR-camera fusion rather than camera-only detection | 3D object detectors · KITTI-C, car moderate | APcor: 2.68; 2.42; 3.05 -> 67.81; 71.87 | Improved |
| Apply motion-level corruptions | 3D object detectors · KITTI-C | AP | Hurt |
| Apply motion-level corruptions | LiDAR-only and LiDAR-camera fusion detectors · nuScenes-C | mAP | Hurt |
| Apply weather-level corruptions | 3D detectors trained on normal-weather data · KITTI-C | AP | Hurt |
| Apply weather-level corruptions | nuScenes-C detectors · nuScenes-C | mAP | No clear effect |
| Corrupt both LiDAR and camera inputs rather than one modality | Focals Conv · KITTI-C | AP | Hurt |
| Use multi-view detection rather than monocular detection | Camera-only detectors · nuScenes-C | mAP: 57%; 54% -> 46%; 45% | Improved |
| Use BEVFormer rather than DETR3D | Camera-only detectors · nuScenes-C, object-level corruptions | mAP | Improved |
| Use LiDAR-only rather than fusion detection for corruptions affecting both modalities | 3D object detectors · KITTI-C | AP | Mixed |
| Use fusion detection rather than LiDAR-only detection for point-cloud-only corruptions | 3D object detectors · KITTI-C | AP | Improved |
Closest papers by meaning
- Comprehensive Robustness Analysis of LiDAR-based 3D Object Detection in Autonomous Driving (ECCV 2026 · similarity 0.92)
- Benchmarking and Analyzing Point Cloud Classification under Corruptions (ICML 2022 · similarity 0.90)
- A Comprehensive Study of the Robustness for LiDAR-Based 3D Object Detectors Against Adversarial Attacks (IJCV 2023 · similarity 0.89)
- Understanding the Robustness of 3D Object Detection With Bird's-Eye-View Representations in Autonomous Driving (CVPR 2023 · similarity 0.88)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.88)
- MOS: Model Synergy for Test-Time Adaptation on LiDAR-Based 3D Object Detection (ICLR 2025 · 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.
