Research map / 3D vision / Human pose estimation
Stacked Hourglass Networks for Human Pose Estimation
Alejandro Newell, Kaiyu Yang, Jia Deng
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 |
|---|---|---|---|
| increase stacks from 2 to 4 hourglasses at equal parameters (ablation) | Stacked Hourglass · MPII validation | accuracy: 87.4% -> 87.8% | Improved |
| increase stacks from 4 to 8 hourglasses at equal parameters (ablation) | Stacked Hourglass · MPII validation | accuracy: 87.8% -> 88.1% | Improved |
| increase stacks from 2 to 8 on halfway intermediate predictions (ablation) | Stacked Hourglass · MPII validation | intermediate accuracy: 84.6% -> 86.5% -> 87.1% | Improved |
| stack two hourglasses instead of a single hourglass with same layers (ablation) | Hourglass · MPII validation training curve | validation accuracy | Improved |
| add intermediate supervision to a single hourglass (ablation) | Hourglass · MPII validation training curve | validation accuracy | Improved |
| combine stacking with intermediate supervision (ablation) | Hourglass · MPII validation training curve | validation accuracy | Improved |
| average heatmaps of original and flipped input at test time (ablation) | Stacked Hourglass · MPII validation | accuracy: 1% average improvement | Improved |
| replace standard large-filter convolutions with residual or Inception modules (ablation) | Hourglass layers · MPII | performance | Improved |
| further explore layer modifications beyond residual modules (ablation) | Hourglass layers · MPII | performance | No clear effect |
| evaluate only visible wrists rather than all wrists | Stacked Hourglass · MPII validation wrist | accuracy: 85.5% -> 93.6% | Improved |
| evaluate only occluded wrists rather than all wrists | Stacked Hourglass · MPII validation wrist | accuracy: 85.5% -> 61.1% | Hurt |
| evaluate only visible elbows rather than all elbows | Stacked Hourglass · MPII validation elbow | accuracy: 90.5% -> 95.1% | Improved |
Closest papers by meaning
- Graph Stacked Hourglass Networks for 3D Human Pose Estimation (CVPR 2021 · similarity 0.93)
- Multi-Context Attention for Human Pose Estimation (CVPR 2017 · similarity 0.92)
- Adversarial PoseNet: A Structure-Aware Convolutional Network for Human Pose Estimation (ICCV 2017 · similarity 0.89)
- Multi-Scale Structure-Aware Network for Human Pose Estimation (ECCV 2018 · similarity 0.88)
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.88)
- Thin-Slicing Network: A Deep Structured Model for Pose Estimation in Videos (CVPR 2017 · similarity 0.88)
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.
