Research map / Label-efficient and robust learning / Recognition and long-tail
Self-Correctable and Adaptable Inference for Generalizable Human Pose Estimation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SCAI to HRNet HR48 | HRNet HR48 · COCO test-dev | AP: 75.5 -> 80.6 | Helped |
| Add SCAI to the strongest listed baseline | COCO test-dev · COCO test-dev | AP: +1.4 | Helped |
| Add SCAI to the strongest listed baseline | CrowdPose test-dev · CrowdPose test-dev | AP: +0.9 | Helped |
| Add a correction network (ablation) | Baseline HRNet HR48 · COCO test-dev | AP: 75.5 -> 78.5 | Helped |
| Add self-referential error feedback (ablation) | Baseline + SCI with correction network · COCO test-dev | AP: 78.5 -> 78.9 | Helped |
| Add joint training of the correction network and FFN (ablation) | Baseline + SCI with self-referential error · COCO test-dev | AP: 78.9 -> 79.8 | Helped |
| Add self-adaptable inference at test time (ablation) | Baseline + SCI · COCO test-dev | AP: 79.8 -> 80.6 | Helped |
| Set λ to 0.30 instead of 0.45 (ablation) | SCAI (a=0.85, b=0.65) · COCO test-dev | AP: 80.6 -> 80.2 | Hurt |
| Set λ to 0.60 instead of 0.45 (ablation) | SCAI (a=0.85, b=0.65) · COCO test-dev | AP: 80.6 -> 80.4 | Hurt |
| Set a to 1.00 instead of 0.85 (ablation) | SCAI (b=0.65, λ=0.45) · COCO test-dev | AP: 80.6 -> 80.5 | Hurt |
| Set b to 0.40 instead of 0.65 (ablation) | SCAI (a=0.85, λ=0.45) · COCO test-dev | AP: 80.6 -> 80.4 | Hurt |
Closest papers by meaning
- PoseFix: Model-Agnostic General Human Pose Refinement Network (CVPR 2019 · similarity 0.88)
- Fast Adaptation for Human Pose Estimation via Meta-Optimization (CVPR 2024 · similarity 0.87)
- SIMPLE: SIngle-network with Mimicking and Point Learning for Bottom-up Human Pose Estimation (AAAI 2021 · similarity 0.87)
- Human Pose Estimation With Iterative Error Feedback (CVPR 2016 · similarity 0.87)
- On the Calibration of Human Pose Estimation (ICML 2024 · similarity 0.85)
- A Dual-Augmentor Framework for Domain Generalization in 3D Human Pose Estimation (CVPR 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.