Research map / 3D vision / Human pose estimation
Poseur: Direct Human Pose Regression with Transformers
Weian Mao, Yongtao Ge, Chunhua Shen, Zhi Tian, Xinlong Wang, Zhibin Wang, Anton van den Hengel
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 |
|---|---|---|---|
| initialize keypoint queries from coarse proposal instead of Deformable DETR reference points (ablation) | Poseur · COCO val | AP: 72.3 -> 72.9 | Improved |
| add noisy reference point sampling during training (ablation) | Poseur · COCO val | AP: 73.7 -> 74.3 | Improved |
| use proposed uncertainty score instead of RLE score (ablation) | Poseur · COCO val | AP: 73.6 -> 74.7 | Improved |
| add proposed uncertainty estimation re-scoring (ablation) | Poseur · COCO val | AP: +4.7 AP | Improved |
| feed more backbone feature levels (1 to 4) (ablation) | Poseur · COCO val | AP: 73.7 -> 74.7 | Improved |
| increase decoder layers from 3 to 4 (ablation) | Poseur · COCO val | AP: 74.7 -> 75.3 | Improved |
| increase decoder layers from 4 to 6 (ablation) | Poseur · COCO val | AP: 75.3 -> 75.4 | No clear effect |
| replace SimpleBaseline heatmap with Poseur at 64x64 input | ResNet-50 · COCO val | AP: 31.4 -> 47.9 | Improved |
| replace SimpleBaseline heatmap with Poseur on MobileNet-V2 | MobileNet-V2 · COCO val | AP: 65.9 -> 71.9 | Improved |
| replace heatmap keypoint head in Mask R-CNN with Poseur | Mask R-CNN ResNet-101 · COCO val | AP: 66.0 -> 68.6 | Improved |
| replace SimpleBaseline with Poseur on ResNet-50 | ResNet-50 · COCO val | AP: 72.4 -> 75.4 | Improved |
Later papers that compared with this method
6 later accepted papers report Poseur as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Video-Based Human Pose Regression via Decoupled Space-Time Aggregation (CVPR 2024 · better on 3 of 3 shared results)
- DiffusionRegPose: Enhancing Multi-Person Pose Estimation using a Diffusion-Based End-to-End Regression Approach (CVPR 2024 · better on 0 of 1 shared result)
- On the Calibration of Human Pose Estimation (ICML 2024 · better on 1 of 1 shared result)
- Human Pose As Compositional Tokens (CVPR 2023 · better on 2 of 2 shared results)
- Group Pose: A Simple Baseline for End-to-End Multi-Person Pose Estimation (ICCV 2023 · better on 1 of 1 shared result)
- Bias-Compensated Integral Regression for Human Pose Estimation (TPAMI 2023 · better on 0 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
- Transformers With Joint Tokens and Local-Global Attention for Efficient Human Pose Estimation (TPAMI 2026 · similarity 0.89)
- Learning Temporal Pose Estimation from Sparsely-Labeled Videos (NeurIPS 2019 · similarity 0.89)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.89)
- Human Pose Estimation via Convolutional Part Heatmap Regression (ECCV 2016 · similarity 0.88)
- Pose Recognition With Cascade Transformers (CVPR 2021 · similarity 0.88)
- LCR-Net: Localization-Classification-Regression for Human Pose (CVPR 2017 · 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.
