Research map / 3D vision / Human pose estimation
Group Pose: A Simple Baseline for End-to-End Multi-Person Pose Estimation
Huan Liu, Qiang Chen, Zichang Tan, Jiang-Jiang Liu, Jian Wang, Xiangbo Su, Xiaolong Li, Kun Yao, Junyu Han, Errui Ding, Yao Zhao, Jingdong Wang
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 standard self-attention with group self-attentions (within-instance then same-type across-instance) (ablation) | GroupPose · MS COCO val2017 | AP: 69.4 -> 72.0 | Improved |
| use group self-attentions instead of standard self-attention with attention mask (ablation) | GroupPose · MS COCO val2017 | AP: 70.7 -> 72.0 | Improved |
| add keypoint queries to instance queries (ablation) | GroupPose · MS COCO val2017 | AP: 64.5 -> 72.0 | Improved |
| add instance queries for classification (ablation) | GroupPose · MS COCO val2017 | AP: 71.2 -> 72.0 | Improved |
| add within-instance self-attention (ablation) | GroupPose · MS COCO val2017 | AP: 67.4 -> 72.0 | Improved |
| add same-type across-instance self-attention (ablation) | GroupPose · MS COCO val2017 | AP: 66.3 -> 72.0 | Improved |
| use separate parameters for the two group self-attentions (ablation) | GroupPose · MS COCO val2017 | AP: 71.5 -> 72.0 | Improved |
| increase number of instances from 100 to 200 (ablation) | GroupPose · MS COCO val2017 | AP: 72.0 -> 72.0 | No clear effect |
| increase number of instances from 50 to 100 (ablation) | GroupPose · MS COCO val2017 | AP: 71.4 -> 72.0 | Improved |
| add human detection decoder layers to GroupPose (ablation) | GroupPose · MS COCO val2017 | AP: 72.0 -> 72.2 | Improved |
| GroupPose with simple decoder, no box supervision, vs ED-Pose with box supervision | ED-Pose · MS COCO val2017 (ResNet-50) | AP: 71.6 -> 72.0 | Improved |
| GroupPose vs Poseur top-down on COCO val2017 ResNet-50 | Poseur · MS COCO val2017 | AP: 74.2 -> 72.0 | Hurt |
Later papers that compared with this method
4 later accepted papers report GroupPose as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- PoseImageNet: Pose Estimation for Extensive Classes Based on Rich Structure Prototypes (ECCV 2026 · better on 2 of 2 shared results)
- DETRPose: Real-Time End-to-End Multi-Person Pose Estimation via Modified Transformer Decoder and Novel Denoising Keypoints (ECCV 2026 · better on 1 of 2 shared results)
- DiffusionRegPose: Enhancing Multi-Person Pose Estimation using a Diffusion-Based End-to-End Regression Approach (CVPR 2024 · better on 1 of 1 shared result)
- Referring Human Pose and Mask Estimation In the Wild (NeurIPS 2024 · 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 1 of 1 on most shared results.
- Poseur — Poseur: Direct Human Pose Regression with Transformers (ECCV 2022) · 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
- Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation (AAAI 2020 · similarity 0.93)
- Explicit Box Detection Unifies End-to-End Multi-Person Pose Estimation (ICLR 2023 · similarity 0.92)
- The Center of Attention: Center-Keypoint Grouping via Attention for Multi-Person Pose Estimation (ICCV 2021 · similarity 0.91)
- DeeperCut: A Deeper, Stronger, and Faster Multi-person Pose Estimation Model (ECCV 2016 · similarity 0.91)
- DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation (CVPR 2016 · similarity 0.91)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · 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.
