Research map / Detection and segmentation / 3D detection and driving
Unsupervised Domain Adaptive 3D Detection With Multi-Level Consistency
Zhipeng Luo, Zhongang Cai, Changqing Zhou, Gongjie Zhang, Haiyu Zhao, Shuai Yi, Shijian Lu, Hongsheng Li, Shanghang Zhang, Ziwei Liu
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 point-level box consistency loss (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 18.61 -> 30.34 | Improved |
| add point-level class consistency loss (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 18.61 -> 20.34 | Improved |
| add instance-level box consistency loss (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 18.61 -> 33.21 | Improved |
| use all four consistency losses at two levels (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 34.95 -> 38.21 | Improved |
| use replicated inputs for matching instead of nearest-point (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 2.93 -> 38.21 | Improved |
| use replicated proposals instead of max-IoU box matching (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 26.95 -> 38.21 | Improved |
| enable neural statistics-level consistency (teacher uses student running BN stats) (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 2.79 -> 38.21 | Improved |
| enable neural statistics consistency instead of separate BN (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 29.88 -> 38.21 | Improved |
| enable EMA update of teacher (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 8.95 -> 38.21 | Improved |
| add input augmentation (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 34.34 -> 36.23 | Improved |
| add both input and RoI augmentation (ablation) | MLC-Net · KITTI->Waymo | AP/L1: 34.34 -> 38.21 | Improved |
| MLC-Net versus Statistical Normalization with target size statistics | PointRCNN · KITTI->Waymo | AP/L1: 30.69 -> 38.21 | Improved |
Closest papers by meaning
- Revisiting Domain-Adaptive 3D Object Detection by Reliable, Diverse and Class-balanced Pseudo-Labeling (ICCV 2023 · similarity 0.92)
- Unsupervised Domain Adaptation for Monocular 3D Object Detection via Self-Training (ECCV 2022 · similarity 0.91)
- Domain Adaptive Object Detection via Asymmetric Tri-way Faster-RCNN (ECCV 2020 · similarity 0.91)
- Unified Domain Generalization and Adaptation for Multi-View 3D Object Detection (NeurIPS 2024 · similarity 0.90)
- Pseudo Label Refinery for Unsupervised Domain Adaptation on Cross-dataset 3D Object Detection (CVPR 2024 · similarity 0.89)
- Geometry-Guided Domain Generalization for Monocular 3D Object Detection (AAAI 2024 · 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.
