AI Paper Atlas

Research map / 3D vision / Human pose estimation

Transformers With Joint Tokens and Local-Global Attention for Efficient Human Pose Estimation

Kaleab A. Kinfu, René Vidal

TPAMI 2026 · 3D vision › Human pose estimation · approach: Heatmap and coordinate pose encoding · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add unified skeletal representation and multi-dataset training (ablation)UniTransPose-B/PS · OCHuman testmAP: 60.7 -> 93.5Improved
add unified skeletal representation and multi-dataset training (ablation)UniTransPose-B/PS · CrowdPose testmAP: 69.9 -> 78.2Improved
add unified skeletal representation and multi-dataset training (ablation)UniTransPose-B/PS · MS-COCO valmAP: 76.1 -> 78.0Improved
add JAGL global attention over local cross-window attention (ablation)UniTransPose-B/PS · MS-COCO valmAP: 77.1 -> 78.0Improved
add JAGL global attention over local shifted-window attention (ablation)UniTransPose-B/PS · OCHuman testmAP: 89.6 -> 93.5Improved
use cross-window instead of shifted-window local attention (ablation)UniTransPose-B/PS · OCHuman testmAP: 89.6 -> 88.3Hurt
use pixel-shuffle decoder instead of classical CNN heat-map decoderUniTransPose-S · MS-COCO valmAP / FLOPs: 75.9 -> 76.2 mAP, 23.5G -> 10.0G FLOPsImproved
use only last feature map instead of hierarchical mapsUniTransPose-S/PS · MS-COCO valmAP: 76.2 -> 73.4Hurt
use key-point regression decoder instead of heat-map decoderUniTransPose/PS · MS-COCO valAP: 78.0 -> 68.0Hurt
add joint-token-based patch selectionEViTPose-B · MS-COCO valmAP / FLOPs: 77.6 -> 76.5 mAP, 19.8G -> 13.7G FLOPsMixed
add joint-token patch selection in place of neighboring selectionEViTPose-B · MS-COCO valmAP: 74.1 -> 76.5Improved
use hierarchical UniTransPose over ViTPose-HUniTransPose-B/PS (384x288) · MS-COCO valmAP: 79.5 -> 79.3Mixed

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