AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Prior-guided Source-free Domain Adaptation for Human Pose Estimation

Dripta S. Raychaudhuri, Calvin-Khang Ta, Arindam Dutta, Rohit Lal, Amit K. Roy-Chowdhury

ICCV 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Geometric alignment and synthetic-to-real adaptation · 21 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 prediction-space pseudo-label self-training (L_out) with Mean Teacher (ablation)source-only pose model · SURREAL->Human3.6MPCK@0.05 avg: 67.3 -> 74.1Improved
add feature-space Barlow Twins consistency (L_feat) (ablation)POST · SURREAL->Human3.6MPCK@0.05 avg: 74.1 -> 75.5Improved
add human pose prior regularization (L_prior) (ablation)POST · SURREAL->Human3.6MPCK@0.05 avg: 75.5 -> 76.7Improved
train pose prior on Human3.6M poses instead of SURREAL poses (ablation)POST · SURREAL->Human3.6MPCK@0.05 avg: 76.7 -> 76.2No clear effect
raise pseudo-label threshold ratio from 0.5 to 0.9 (ablation)POST · SURREAL->Human3.6MPCK@0.05 avg: 76.7 -> 73.8Hurt
lower pseudo-label threshold ratio from 0.5 to 0.1 (ablation)POST · SURREAL->Human3.6MPCK@0.05 avg: 76.7 -> 76.4No clear effect
apply POST source-free adaptationsource pose model · SURREAL->LSPPCK@0.05 avg: 63.9 -> 80.3Improved
compare source-free POST with UDAPE that uses source dataPOST · SURREAL->LSPPCK@0.05 avg: 80.3 vs 82.0 (UDAPE); 76.7 vs 79.0 on Human3.6MMixed

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