Research map / Detection and segmentation / 3D detection and driving
PETR: Position Embedding Transformation for Multi-View 3D Object Detection
Yingfei Liu, Tiancai Wang, Xiangyu Zhang, Jian Sun
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 3D position embedding to 2D PE (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.208 -> 0.356 | Improved |
| add multi-view PE to 2D PE (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.208 -> 0.224 | Improved |
| add 2D PE and multi-view PE on top of 3D PE (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.356 -> 0.359 | No clear effect |
| add MLP on 3D coordinates (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.311 -> 0.359 | Improved |
| use 3x3 convs instead of 1x1 in PE network (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.359 -> 0.017 | Hurt |
| fuse by concatenation instead of addition (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.359 -> 0.358 | No clear effect |
| fuse by multiplication instead of addition (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.359 -> 0.357 | Hurt |
| use learned 3D anchor points instead of fixed BEV anchors (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.337 -> 0.359 | Improved |
| use learned 3D anchor points instead of fixed 3D anchors (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.343 -> 0.359 | Improved |
| use anchor points instead of DETR learnable queries (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: - -> 0.359 | Improved |
| increase anchor points from 600 to 1500 (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.339 -> 0.359 | Improved |
| linear-increasing instead of uniform depth discretization (ablation) | PETR · nuScenes val (ablation, ResNet-50 C5) | NDS: 0.358 -> 0.359 | No clear effect |
Later papers that compared with this method
20 later accepted papers report PETRv2 as a baseline in their results tables; 17 of them report a better number on most of the results they share with it.
- CycleBEV: Regularizing View Transformation Networks via View Cycle Consistency for Bird's-Eye-View Semantic Segmentation (CVPR 2026 · better on 3 of 3 shared results)
- SemLT3D: Semantic-Guided Expert Distillation for Camera-only Long-Tailed 3D Object Detection (CVPR 2026 · better on 2 of 2 shared results)
- Object-Scene-Camera Decomposition and Recomposition for Data Efficient Monocular 3D Object Detection (IJCV 2026 · better on 2 of 2 shared results)
- Exploring Surround-View Fisheye Camera 3D Object Detection (AAAI 2026 · better on 0 of 1 shared result)
- Refine3D: Scene-Adaptive Reference Point Refinement for Sparse 3D Object Detection (AAAI 2026 · better on 1 of 1 shared result)
- Toward Unified 3D Object Detection via Algorithm and Data Unification (TPAMI 2025 · better on 2 of 2 shared results)
- RepVF: A Unified Vector Fields Representation for Multi-task 3D Perception (ECCV 2024 · better on 1 of 4 shared results)
- EASE-DETR: Easing the Competition among Object Queries (CVPR 2024 · better on 2 of 2 shared results)
12 more not listed.
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
- PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images (ICCV 2023 · similarity 0.92)
- FQ-PETR: Fully Quantized Position Embedding Transformation for Multi-View 3D Object Detection (AAAI 2026 · similarity 0.91)
- OPEN: Object-wise Position Embedding for Multi-view 3D Object Detection (ECCV 2024 · similarity 0.85)
- V-DETR: DETR with Vertex Relative Position Encoding for 3D Object Detection (ICLR 2024 · similarity 0.84)
- Towards Generalizable Multi-Camera 3D Object Detection via Perspective Rendering (AAAI 2025 · similarity 0.84)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.83)
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.
