Research map / 3D vision / Human pose estimation
Monocular 3D Human Pose Estimation by Predicting Depth on Joints
Bruce Xiaohan Nie, Ping Wei, Song-Chun Zhu
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 patch-LSTM local image evidence to skeleton-LSTM (ablation) | our(s) · Human3.6M (Protocol 1) | MPJPE (mm): 84.9 -> 79.5 | Improved |
| add skeleton-LSTM global 2D skeleton features to patch-LSTM (ablation) | our(p) · Human3.6M (Protocol 1) | MPJPE (mm): 108.4 -> 79.5 | Improved |
| add patch-LSTM to skeleton-LSTM, Protocol-1 with ground-truth 2D poses (ablation) | our(s) · Human3.6M (Protocol 1, ground-truth 2D poses) | MPJPE (mm): 46.3 -> 42.9 | Improved |
| add patch-LSTM to skeleton-LSTM on HHOI PullUp (ablation) | our(s) · HHOI PullUp | MPJPE (mm): 139.8 -> 124.8 | Improved |
| add patch-LSTM to skeleton-LSTM on HHOI ShakeHands (ablation) | our(s) · HHOI ShakeHands | MPJPE (mm): 113.1 -> 118.6 | Hurt |
| allow scale transformation at evaluation (ablation) | our(s+p) · Human3.6M (Protocol 1) | MPJPE (mm): 79.5 -> 74.0 | Improved |
| use tree-structured LSTM instead of sequential chain on skeleton features (ablation) | ske-seq · Human3.6M (Protocol 1, model structure) | MPJPE (mm): 89.0 -> 84.9 | Improved |
| use tree-structured LSTM instead of sequential chain on patch features (ablation) | patch-seq · Human3.6M (Protocol 1, model structure) | MPJPE (mm): 118.6 -> 108.4 | Improved |
| add LSTM joint relation modeling to 2D pose encoding on skeleton features (ablation) | ske-encoding · Human3.6M (Protocol 1, model structure) | MPJPE (mm): 113.0 -> 84.9 | Improved |
| use LSTM on body part patches instead of VGG on whole person (ablation) | whole-vgg · Human3.6M (Protocol 1, model structure) | MPJPE (mm): 169.8 -> 108.4 | Improved |
| skeleton-only model vs prior state of the art Yasin et al. | our(s) · Human3.6M (Protocol 1) | MPJPE (mm): 110.2 -> 84.9 | Improved |
| final model vs Zhou et al. despite their temporal information | our(s+p) · Human3.6M (Protocol 2) | MPJPE (mm): 113.0 -> 97.5 | Improved |
Closest papers by meaning
- Sparseness Meets Deepness: 3D Human Pose Estimation From Monocular Video (CVPR 2016 · similarity 0.91)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.91)
- Propagating LSTM: 3D Pose Estimation based on Joint Interdependency (ECCV 2018 · similarity 0.90)
- Generating Multiple Hypotheses for 3D Human Pose Estimation With Mixture Density Network (CVPR 2019 · similarity 0.89)
- Exploiting temporal information for 3D human pose estimation (ECCV 2018 · similarity 0.89)
- From Human Pose Similarity Metric to 3D Human Pose Estimator: Temporal Propagating LSTM Networks (TPAMI 2022 · similarity 0.89)
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.
