Research map / Label-efficient and robust learning / Domain adaptation and generalization
Source-free Domain Adaptive Human Pose Estimation
Qucheng Peng, Ce Zheng, Chen Chen
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 the target-relevant (TR) module to MMT (ablation) | MMT · RHD -> H3D | PCK@0.05: 75.3 -> 79.1 | Improved |
| add the source-protect (SP) module in place of MMT (ablation) | MMT + TR · RHD -> H3D | PCK@0.05: 79.1 -> 82.2 | Improved |
| add pose-specific contrastive loss L_cst (ablation) | baseline (L_ft + L_con) · RHD -> H3D | PCK@0.05: 79.4 -> 80.9 | Improved |
| add residual loss L_res (ablation) | baseline (L_ft + L_con) · SURREAL -> Human3.6M | PCK@0.05: 75.8 -> 76.6 | Improved |
| add pose-specific information maximization L_im (ablation) | baseline (L_ft + L_con) · SURREAL -> Human3.6M | PCK@0.05: 75.8 -> 76.9 | Improved |
| use heatmap-based contrastive learning instead of no contrastive term (ablation) | MMT · RHD -> H3D | PCK@0.05: 75.3 -> 74.2 | Hurt |
| use vector-based spatial space for contrastive learning (ablation) | MMT · RHD -> H3D | PCK@0.05: 75.3 -> 78.7 | Improved |
| use vector-based instead of heatmap-based information maximization (ablation) | MMT · RHD -> H3D | PCK@0.05: 75.9 -> 77.5 | Improved |
| vary alpha, beta, gamma hyperparameters (ablation) | Ours · RHD -> H3D | PCK@0.05: decreases limited to 0.2% | No clear effect |
| compare source-free method against UniFrame that uses source data | Ours · SURREAL -> Human3.6M | PCK@0.05: 79.0 -> 78.7 | Hurt |
Closest papers by meaning
- Lifelong Domain Adaptive 3D Human Pose Estimation (AAAI 2026 · similarity 0.91)
- Prior-guided Source-free Domain Adaptation for Human Pose Estimation (ICCV 2023 · similarity 0.91)
- Fast Adaptation for Human Pose Estimation via Meta-Optimization (CVPR 2024 · similarity 0.88)
- A Dual-Augmentor Framework for Domain Generalization in 3D Human Pose Estimation (CVPR 2024 · similarity 0.87)
- Source-Free and Image-Only Unsupervised Domain Adaptation for Category Level Object Pose Estimation (ICLR 2024 · similarity 0.86)
- PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF (AAAI 2024 · similarity 0.85)
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.
