Research map / 3D vision / Human pose estimation
Fast Human Pose Estimation
Feng Zhang, Xiatian Zhu, Mao Ye
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add pose knowledge distillation to the lightweight Hourglass (ablation) | FPD · MPII validation | Mean PCKh@0.5: 90.1 -> 90.9 | Improved |
| reduce Hourglass stages from 8 to 4 (ablation) | Hourglass · MPII validation | Mean PCKh@0.5: 91.9 -> 91.4 | Hurt |
| reduce Hourglass stages from 8 to 2 (ablation) | Hourglass · MPII validation | Mean PCKh@0.5: 91.9 -> 90.5 | Hurt |
| reduce Hourglass stages from 8 to 1 (ablation) | Hourglass · MPII validation | Mean PCKh@0.5: 91.9 -> 86.4 | Hurt |
| reduce channels from 256 to 128 in a 4-stage Hourglass (ablation) | Hourglass · MPII validation | Mean PCKh@0.5: 91.4 -> 90.1 | Hurt |
| reduce channels from 128 to 64 (ablation) | 4-stage Hourglass · MPII validation | Mean PCKh@0.5: 90.1 -> 87.9 | Hurt |
| reduce channels from 128 to 32 (ablation) | 4-stage Hourglass · MPII validation | Mean PCKh@0.5: 90.1 -> 83.4 | Hurt |
| replace MSE with cross-entropy for pose distillation (ablation) | FPD · MPII validation | Mean PCKh@0.5: 90.9 -> 90.7 | Hurt |
| set distillation weight alpha to 0 instead of 0.5 (ablation) | FPD · MPII validation | Mean PCKh@0.5: 90.9 -> 90.1 | Hurt |
| vary alpha from 0.05 to 0.99 around the 0.5 setting (ablation) | FPD · MPII validation | Mean PCKh@0.5: 90.8; 90.8; 90.9; 90.7; 90.7 | No clear effect |
| use the lightweight model instead of the strongest listed baseline | FPD vs. Nie et al. · MPII test | Mean PCKh@0.5: 92.4 -> 91.1 | Hurt |
| use FPD instead of the most efficient listed competitor | FPD vs. Sekii · MPII test | Mean PCKh@0.5: 88.1 -> 91.1 | Improved |
Closest papers by meaning
- Online Knowledge Distillation for Efficient Pose Estimation (ICCV 2021 · similarity 0.89)
- Dynamic Kernel Distillation for Efficient Pose Estimation in Videos (ICCV 2019 · similarity 0.89)
- Learning Pose Grammar to Encode Human Body Configuration for 3D Pose Estimation (AAAI 2018 · similarity 0.88)
- DistilPose: Tokenized Pose Regression With Heatmap Distillation (CVPR 2023 · similarity 0.86)
- SDPose: Tokenized Pose Estimation via Circulation-Guide Self-Distillation (CVPR 2024 · similarity 0.86)
- Fast Adaptation for Human Pose Estimation via Meta-Optimization (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.
