Research map / Point clouds / Segmentation and detection
CenterFormer: Center-based Transformer for 3D Object Detection
Zixiang Zhou, Xiangchen Zhao, Yu Wang, Panqu Wang, Hassan Foroosh
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 multi-scale center proposal network (ablation) | center-based baseline · WOD val (single frame, mean L2 mAPH) | mAPH L2: 65.9 -> 66.1 | No clear effect |
| add transformer self-attention layer (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 66.1 -> 66.8 | Improved |
| add both self- and cross-attention layers (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 66.1 -> 67.0 | Improved |
| add corner auxiliary supervision (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 67.0 -> 67.1 | No clear effect |
| add IoU rectification (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 67.1 -> 68.7 | Improved |
| add fade augmentation (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 68.7 -> 69.1 | Improved |
| replace cross-attention with deformable cross-attention (ablation) | CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 67.1 -> 67.3 | Improved |
| add IoU rectification to deformable model (ablation) | Deformable CenterFormer · WOD val (single frame, mean L2 mAPH) | mAPH L2: 67.3 -> 68.3 | Improved |
| use center feature as query instead of learnable embedding (ablation) | CenterFormer · WOD val | mAPH: +1.5% mAPH | Improved |
| use center-based training instead of DETR set matching (ablation) | CenterFormer · WOD val | mAPH: 46.3 | Improved |
| add position embedding (ablation) | CenterFormer · WOD val (20% train) | mAPH L2: 59.5 | Improved |
| increase number of input frames from 1 to 8 | Deformable CenterFormer · WOD val | mean L2 mAPH: 69.0 -> 73.7 | Improved |
Later papers that compared with this method
7 later accepted papers report CenterFormer as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.
- PTNET: A PROPOSAL-CENTRIC TRANSFORMER NET- WORK FOR 3D OBJECT DETECTION (ICLR 2026 · better on 2 of 2 shared results)
- Clusterformer: Cluster-based Transformer for 3D Object Detection in Point Clouds (ICCV 2023 · better on 3 of 3 shared results)
- FlatFormer: Flattened Window Attention for Efficient Point Cloud Transformer (CVPR 2023 · better on 2 of 2 shared results)
- MSF: Motion-Guided Sequential Fusion for Efficient 3D Object Detection From Point Cloud Sequences (CVPR 2023 · better on 2 of 2 shared results)
- DSVT: Dynamic Sparse Voxel Transformer With Rotated Sets (CVPR 2023 · better on 1 of 1 shared result)
- HEDNet: A Hierarchical Encoder-Decoder Network for 3D Object Detection in Point Clouds (NeurIPS 2023 · better on 1 of 1 shared result)
- DetZero: Rethinking Offboard 3D Object Detection with Long-term Sequential Point Clouds (ICCV 2023 · better on 1 of 1 shared result)
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.
- LiDAR R-CNN — LiDAR R-CNN: An Efficient and Universal 3D Object Detector (CVPR 2021) · better on 1 of 1 shared result
- PV-RCNN++ — PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (CVPR 2020) · 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
- 3D Object Detection With Pointformer (CVPR 2021 · similarity 0.89)
- Finding Your (3D) Center: 3D Object Detection Using a Learned Loss (ECCV 2020 · similarity 0.89)
- SEFormer: Structure Embedding Transformer for 3D Object Detection (AAAI 2023 · similarity 0.89)
- An End-to-End Transformer Model for 3D Object Detection (ICCV 2021 · similarity 0.89)
- PTNET: A PROPOSAL-CENTRIC TRANSFORMER NET- WORK FOR 3D OBJECT DETECTION (ICLR 2026 · similarity 0.88)
- GeoFormer: Geometry Point Encoder for 3D Object Detection with Graph-based Transformer (ICCV 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.
