Research map / 3D vision / Human pose estimation
Imitation Learning for Human Pose Prediction
Borui Wang, Ehsan Adeli, Hsu-kuang Chiu, De-An Huang, Juan Carlos Niebles
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 WGAIL-div stage after behavioral cloning (ablation) | BC pose predictor · Human3.6M (avg of walking, eating, smoking, discussion) | mean angle error @1000ms: 1.36 -> 1.32 | Improved |
| add WGAIL-div stage after behavioral cloning (ablation) | BC pose predictor · Human3.6M (avg of walking, eating, smoking, discussion) | mean angle error @80ms: 0.23 -> 0.21 | Improved |
| add BC pre-training before WGAIL-div (ablation) | WGAIL-div pose predictor · Human3.6M (avg of walking, eating, smoking, discussion) | mean angle error @400ms: 0.80 -> 0.75 | Improved |
| add BC pre-training before WGAIL-div (ablation) | WGAIL-div pose predictor · Human3.6M (avg of walking, eating, smoking, discussion) | mean angle error @1000ms: 1.33 -> 1.32 | Improved |
| add BC pre-training before WGAIL-div (ablation) | WGAIL-div pose predictor · Human3.6M (avg of walking, eating, smoking, discussion) | mean angle error @80ms: 0.23 -> 0.21 | Improved |
| replace TP-RNN with imitation learning (BC+WGAIL-div) | TP-RNN · Human3.6M (average over 15 actions, long-term) | mean angle error @1000ms: 1.71 -> 1.65 | Improved |
| replace TP-RNN with imitation learning (BC+WGAIL-div) | TP-RNN · Human3.6M (average over 15 actions, long-term) | mean angle error @80ms: 0.37 -> 0.31 | Improved |
| replace Residual with imitation learning (BC+WGAIL-div) | Residual · Human3.6M (average over 15 actions, long-term) | mean angle error @1000ms: 1.83 -> 1.65 | Improved |
| replace Residual and TP-RNN training with BC+WGAIL-div | Residual and TP-RNN · Human3.6M training time | training time: more than 8 hours -> within 20 minutes | Improved |
Closest papers by meaning
- The Pose Knows: Video Forecasting by Generating Pose Futures (ICCV 2017 · similarity 0.87)
- Deep Reinforcement Learning for Active Human Pose Estimation (AAAI 2020 · similarity 0.87)
- Model-Based Robot Imitation with Future Image Similarity (IJCV 2019 · similarity 0.87)
- Generating Smooth Pose Sequences for Diverse Human Motion Prediction (ICCV 2021 · similarity 0.87)
- FutureHuman3D: Forecasting Complex Long-Term 3D Human Behavior from Video Observations (CVPR 2024 · similarity 0.86)
- SimpliHuMoN: Simplifying Human Motion Prediction (NeurIPS 2026 · 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.
