Research map / 3D vision / Human pose estimation
Differentiable Hierarchical Graph Grouping for Multi-Person Pose Estimation
Sheng Jin, Wentao Liu, Enze Xie, Wenhai Wang, Chen Qian, Wanli Ouyang, Ping Luo
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 end-to-end grouping loss (final macro-node discriminator with clustering) (ablation) | AE baseline · COCO val | AP: 57.6 -> 58.2 | Improved |
| add end-to-end grouping loss (ablation) | AE baseline · COCO val | AR: 62.1 -> 63.4 | Improved |
| add edge discriminator (ablation) | AE baseline · COCO val | AP: 57.6 -> 59.6 | Improved |
| add final macro-node supervision (ablation) | HGG with edge discriminator · COCO val | AP: 59.6 -> 60.1 | Improved |
| add intermediate macro-node supervision (ablation) | HGG · COCO val | AP: 60.1 -> 60.4 | Improved |
| train with grouping losses but keep greedy post-processing at inference (ablation) | AE baseline · COCO val | AP: 57.6 -> 59.6 | Improved |
| use EdgeConv instead of FC for graph interaction (ablation) | HGG · COCO val | AP: 58.3 -> 60.4 | Improved |
| use EdgeConv instead of GAT (ablation) | HGG · COCO val | AP: 59.3 -> 60.4 | Improved |
| use fully-connected graph instead of tree-structured graph (ablation) | HGG · COCO val | mAP: 56.1 -> 60.4 | Improved |
| use denser graph (Tree to Full) on runtime (ablation) | HGG · COCO val | runtime | No clear effect |
| add multi-scale testing (ablation) | HGG · COCO val | AP: 60.4 -> 68.3 | Improved |
| add multi-scale testing (ablation) | AE baseline · COCO val | AP: 57.6 -> 65.6 | Improved |
Closest papers by meaning
- Simple Pose: Rethinking and Improving a Bottom-up Approach for Multi-Person Pose Estimation (AAAI 2020 · similarity 0.90)
- DGCN: Dynamic Graph Convolutional Network for Efficient Multi-Person Pose Estimation (AAAI 2020 · similarity 0.89)
- Graph-Based 3D Multi-Person Pose Estimation Using Multi-View Images (ICCV 2021 · similarity 0.89)
- PoseTrack: Joint Multi-Person Pose Estimation and Tracking (CVPR 2017 · similarity 0.89)
- Monocular 3D Multi-Person Pose Estimation by Integrating Top-Down and Bottom-Up Networks (CVPR 2021 · similarity 0.89)
- Group Pose: A Simple Baseline for End-to-End Multi-Person Pose Estimation (ICCV 2023 · 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.
