Research map / 3D vision / Human pose estimation
Maximum-Margin Structured Learning with Deep Networks for 3D Human Pose Estimation
Sijin Li, Weichen Zhang, Antoni B. Chan
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 |
|---|---|---|---|
| average the top-500 scoring poses instead of taking the max pose (ablation) | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 135.63 -> 121.46 | Improved |
| average top-20 poses instead of max (ablation) | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 135.63 -> 128.11 | Improved |
| increase averaged poses from 500 to 1500 (ablation) | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 121.46 -> 123.04 | No clear effect |
| apply APF to the averaged pose to get a valid pose (ablation) | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 121.46 -> 121.31 | No clear effect |
| add auxiliary pose prediction task (lambda=1) (ablation) | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 145.74 -> 135.63 | Improved |
| use 2-layer pose embedding instead of raw joint coordinates (ablation) | StructNet · Human3.6M test, Walking | MPJPE (mm): 114.24 -> 83.64 | Improved |
| use 1-layer learned embedding instead of fixed random weights (ablation) | StructNet · Human3.6M test, Walking | MPJPE (mm): 87.94 -> 86.48 | Improved |
| use 2-layer instead of 1-layer pose embedding (ablation) | StructNet · Human3.6M test, Walking | MPJPE (mm): 86.48 -> 83.64 | Improved |
| use StructNet-Avg(500)-APF over DconvMP-HML | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 133.54 -> 121.31 | Improved |
| use StructNet-Avg(500) over LinKDE on hidden test set | StructNet · Human3.6M online hidden test | MPJPE (mm): 133.81 -> 122.03 | Improved |
| use score-function poses instead of auxiliary regression poses | StructNet · Human3.6M test (S9,S11), all actions | MPJPE (mm): 133.79 -> 121.46 | Improved |
Closest papers by meaning
- Adversarial PoseNet: A Structure-Aware Convolutional Network for Human Pose Estimation (ICCV 2017 · similarity 0.90)
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.90)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.89)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.89)
- Learning Pose Grammar to Encode Human Body Configuration for 3D Pose Estimation (AAAI 2018 · similarity 0.89)
- 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.
