Research map / 3D vision / Human pose estimation
Propagating LSTM: 3D Pose Estimation based on Joint Interdependency
Kyoungoh Lee, Inwoong Lee, Sanghoon Lee
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 |
|---|---|---|---|
| replace the single LSTM with connected propagating LSTMs (ablation) | 3D pose lifting network (S=1, T=1) · Human3.6M Protocol 1 | MPJPE (mm): 108.5 -> 58.4 | Improved |
| add multi-stage refinement (1 to 3 stages) (ablation) | p-LSTMs · Human3.6M Protocol 1 | MPJPE (mm): 58.4 -> 55.8 | Improved |
| add multi-stage refinement (1 to 3 stages) (ablation) | single LSTM baseline · Human3.6M Protocol 1 | MPJPE (mm): 108.5 -> 104.3 | Improved |
| use 3 input frames instead of 1 (ablation) | p-LSTMs (S=3) · Human3.6M Protocol 1 | MPJPE (mm): 55.8 -> 52.8 | Improved |
| use 3 input frames instead of 1 (ablation) | single LSTM baseline (S=3) · Human3.6M Protocol 1 | MPJPE (mm): 104.3 -> 97.6 | Improved |
| use 10 input frames instead of 3 (ablation) | p-LSTMs (S=3) · Human3.6M Protocol 1 | MPJPE (mm): 52.8 -> 68.8 | Hurt |
| use 5 input frames instead of 3 (ablation) | p-LSTMs (S=3) · Human3.6M Protocol 1 | MPJPE (mm): 52.8 -> 60.2 | Mixed |
| add auxiliary 2D pose in the pose depth cue (replacement instead of elimination and addition) (ablation) | p-LSTMs (S=3, T=1) · Human3.6M Protocol 1 | MPJPE (mm): 87.6 -> 55.8 | Improved |
| use replacement cue instead of concatenation cue (ablation) | p-LSTMs (S=3, T=1) · Human3.6M Protocol 1 | MPJPE (mm): 79.8 -> 55.8 | Improved |
| propagate outward from the body center instead of inward (ablation) | p-LSTMs (S=3, T=1) · Human3.6M Protocol 1 | MPJPE (mm): 83.6 -> 55.8 | Improved |
| p-LSTMs versus prior best single-frame method | p-LSTMs vs Sun ICCV'17 · Human3.6M Protocol 1 | MPJPE (mm): 59.1 -> 55.8 | Improved |
| p-LSTMs versus Martinez with GT 2D input | p-LSTMs vs Martinez ICCV'17 · Human3.6M Protocol 1 (GT 2D) | MPJPE (mm): 45.5 -> 40.9 | Improved |
Closest papers by meaning
- Monocular 3D Human Pose Estimation by Predicting Depth on Joints (ICCV 2017 · similarity 0.90)
- From Human Pose Similarity Metric to 3D Human Pose Estimator: Temporal Propagating LSTM Networks (TPAMI 2022 · similarity 0.89)
- Exploiting temporal information for 3D human pose estimation (ECCV 2018 · similarity 0.88)
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.88)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.88)
- SRNet: Improving Generalization in 3D Human Pose Estimation with a Split-and-Recombine Approach (ECCV 2020 · 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.
