Research map / 3D vision / Human pose estimation
Invariant Teacher and Equivariant Student for Unsupervised 3D Human Pose Estimation
Chenxin Xu, Siheng Chen, Maosen Li, Ya Zhang
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 the rotation-invariant cycle loss to the teacher (ablation) | ITES-T on Human3.6M GT · Human3.6M GT | MPJPE: 153.5 -> 85.6 | Improved |
| add knowledge distillation from the teacher (ablation) | ITES-TS on Human3.6M GT · Human3.6M GT | MPJPE: 192.3 -> 82.4 | Improved |
| add the rotation-equivariant cycle loss to the distilled student (ablation) | ITES-TS on Human3.6M GT · Human3.6M GT | MPJPE: 82.4 -> 77.2 | Improved |
| replace C3DPO teacher with the pose-dictionary teacher (ablation) | ITES teacher with AGCN student on Human3.6M GT · Human3.6M GT | MPJPE: 91.7 -> 77.2 | Improved |
| replace ResFC student with AGCN (ablation) | pose-dictionary teacher on Human3.6M GT · Human3.6M GT | MPJPE: 82.3 -> 77.2 | Improved |
| replace teacher ResFC feature extraction with AGCN (ablation) | ITES-T on Human3.6M GT · Human3.6M GT | MPJPE: 85.9 -> 112.3 | Hurt |
| add nonphysical graph connections to the physical graph (ablation) | ITES student on Human3.6M GT · Human3.6M GT | MPJPE: 81.1 -> 77.2 | Improved |
| add physical graph connections to the nonphysical graph (ablation) | ITES student on Human3.6M GT · Human3.6M GT | MPJPE: 83.3 -> 77.2 | Improved |
| vary pose-dictionary size away from the selected size (ablation) | ITES teacher on Human3.6M · Human3.6M | pose estimation error | Hurt |
| increase teacher reprojection-loss weight from 1 to 5 while keeping invariant-loss weight at 1 (ablation) | ITES-T on Human3.6M GT · Human3.6M GT | MPJPE: 144.3 -> 85.6 | Improved |
| increase teacher reprojection-loss weight from 5 to 10 while keeping invariant-loss weight at 1 (ablation) | ITES-T on Human3.6M GT · Human3.6M GT | MPJPE: 85.6 -> 96.7 | Hurt |
| increase student distillation-loss weight from 1 to 5 while keeping equivariant-loss weight at 1 (ablation) | ITES-TS on Human3.6M GT · Human3.6M GT | MPJPE: 85.2 -> 77.2 | Improved |
Closest papers by meaning
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.90)
- Unsupervised Geometry-Aware Representation for 3D Human Pose Estimation (ECCV 2018 · similarity 0.89)
- Exploiting temporal information for 3D human pose estimation (ECCV 2018 · similarity 0.89)
- Kinematic-Structure-Preserved Representation for Unsupervised 3D Human Pose Estimation (AAAI 2020 · similarity 0.89)
- Weakly-Supervised Discovery of Geometry-Aware Representation for 3D Human Pose Estimation (CVPR 2019 · similarity 0.88)
- Unsupervised Learning of Category-Level 3D Pose from Object-Centric Videos (CVPR 2024 · similarity 0.87)
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.
