Research map / Detection and segmentation / 3D detection and driving
3DPPE: 3D Point Positional Encoding for Transformer-based Multi-Camera 3D Object Detection
Changyong Shu, Jiajun Deng, Fisher Yu, Yifan 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 |
|---|---|---|---|
| replace camera-ray PE with oracle 3D point PE using ground-truth depth (ablation) | PETR-style detector · nuScenes val | NDS: 0.337 -> 0.404 | Improved |
| replace camera-ray PE with LiDAR-ray PE (ablation) | PETR-style detector · nuScenes val | NDS: 0.337 -> 0.338 | No clear effect |
| add smooth-L1 depth supervision (ablation) | 3DPPE · nuScenes val | NDS: 0.343 -> 0.365 | Improved |
| add distribution focal loss on top of smooth-L1 (ablation) | 3DPPE · nuScenes val | NDS: 0.365 -> 0.368 | Improved |
| use depth-guided single 3D point PE instead of camera-ray PE (ablation) | 3DPPE · nuScenes val | NDS: 0.337 -> 0.368 | Improved |
| use top-5 depth bins instead of one predicted depth point (ablation) | 3DPPE · nuScenes val | NDS: 0.368 -> 0.327 | Hurt |
| add depth feature into positional encoding (ablation) | 3DPPE · nuScenes val | NDS: 0.368 -> 0.359 | Hurt |
| share 3D point PE encoder between image features and object queries (ablation) | 3DPPE · nuScenes val | NDS: 0.362 -> 0.368 | Improved |
| add temporal modeling (3DPPE-v2) versus PETR-v2 | 3DPPE · nuScenes val | NDS: 0.502 -> 0.524 | Improved |
| add knowledge distillation from oracle-depth model | 3DPPE · nuScenes val | NDS: 0.440 -> 0.454 | Improved |
| replace PETR camera-ray PE with 3DPPE point PE | PETR · nuScenes test | mAP: 0.441 -> 0.460 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PETR — PETR: Position Embedding Transformation for Multi-View 3D Object Detection (ECCV 2022) · better on 5 of 5 shared results
- PETR-v2 — PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images (ICCV 2023) · 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
- OPEN: Object-wise Position Embedding for Multi-view 3D Object Detection (ECCV 2024 · similarity 0.90)
- Improving 3D Object Detection With Channel-Wise Transformer (ICCV 2021 · similarity 0.89)
- Cubify Anything: Scaling Indoor 3D Object Detection (CVPR 2025 · similarity 0.87)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.87)
- MonoCD: Monocular 3D Object Detection with Complementary Depths (CVPR 2024 · similarity 0.86)
- BEVNeXt: Reviving Dense BEV Frameworks for 3D Object Detection (CVPR 2024 · similarity 0.86)
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.
