AI Paper Atlas

Research map / 3D vision / Human pose estimation

Distribution-Aware Single-Stage Models for Multi-Person 3D Pose Estimation

Zitian Wang, Xuecheng Nie, Xiaochao Qu, Yunpeng Chen, Si Liu

CVPR 2022 · 3D vision › Human pose estimation · approach: Multi-person pose estimation · 47 citations (OpenAlex)

Check 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 changedWhereMeasuredEffect
Use root joint rather than bounding-box center (ablation)DAS · CMU PanopticMPJPE: 65.3 -> 62.5Improved
Add recursive update (ablation)DAS with root center · CMU PanopticMPJPE: 62.5 -> 57.6Improved
Add MLE for the learned joint-location distribution (ablation)DAS with root center · CMU PanopticMPJPE: 62.5 -> 58.8Improved
Combine recursive update with MLE (ablation)DAS with root center · CMU PanopticMPJPE: 62.5 -> 56.3Improved
Use a larger backbone (ablation)DAS with recursive update and MLE · CMU PanopticMPJPE: 56.3 -> 54.4Improved
Add recursive update (ablation)DAS · CMU PanopticMPJPE: 62.5 -> 58.6Improved
Add multi-source update (ablation)DAS with one recursive update layer · CMU PanopticMPJPE: 58.6 -> 58.2Improved
Increase recursive update depth from one to two layers (ablation)DAS with multi-source update · CMU PanopticMPJPE: 58.2 -> 57.9Improved
Increase recursive update depth from two to three layers (ablation)DAS with multi-source update · CMU PanopticMPJPE: 57.9 -> 57.6Improved
Use single-stage DAS instead of two-stage top-down methodsMulti-person 3D pose estimation · MuPoTS-3DPCK_rel: 82.7 vs 83.7*Mixed
Use single-stage DAS instead of bottom-up methodsMulti-person 3D pose estimation · MuPoTS-3DPCK_rel: 82.7 vs 73.5Improved
Use single-stage DAS instead of the strongest listed CMU Panoptic baselineMulti-person 3D pose estimation · CMU PanopticMPJPE: 55.1* -> 53.8Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy