Research map / 3D vision / Human pose estimation
Direct Prediction of 3D Body Poses From Motion Compensated Sequences
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use motion-compensated spatiotemporal volumes rather than unaligned volumes (ablation) | RSTV pose regression · Human3.6m Greeting | mean Euclidean joint position error (mm): 144.48 -> 127.12 | Helped |
| Use motion-compensated spatiotemporal volumes rather than unaligned volumes (ablation) | RSTV pose regression · Human3.6m Walking Dog | mean Euclidean joint position error (mm): 138.66 -> 126.29 | Helped |
| Use CNN-based motion compensation instead of optical-flow compensation (ablation) | RSTV pose regression · Human3.6m Greeting | mean Euclidean joint position error (mm): 140.97 -> 127.12 | Helped |
| Use CNN-based motion compensation instead of optical-flow compensation (ablation) | RSTV pose regression · Human3.6m Walking Dog | mean Euclidean joint position error (mm): 134.98 -> 126.29 | Helped |
| Use a 24-frame rather than 12-frame temporal window (ablation) | RSTV+DN · Human3.6m Walking | mean Euclidean joint position error (mm): 58.78 -> 55.07 | Helped |
| Use a 36-frame rather than 12-frame temporal window (ablation) | RSTV+DN · Human3.6m Walking | mean Euclidean joint position error (mm): 58.78 -> 53.68 | Helped |
| Use a 48-frame rather than 12-frame temporal window (ablation) | RSTV+DN · Human3.6m Eating | mean Euclidean joint position error (mm): 93.97 -> 85.36 | Helped |
| Use a 36-frame rather than 12-frame temporal window (ablation) | RSTV+DN · Human3.6m Eating | mean Euclidean joint position error (mm): 93.97 -> 87.23 | Helped |
| Use a 48-frame rather than 36-frame temporal window (ablation) | RSTV+DN · Human3.6m Walking | mean Euclidean joint position error (mm): 53.68 -> 54.36 | Hurt |
| Regress from rectified spatiotemporal volumes rather than single-frame features | 3D pose estimation · Human3.6m | mean Euclidean joint position error (mm): 162.14 (99.38) -> 124.97 (57.72) | Helped |
| Regress from rectified spatiotemporal volumes rather than single-frame features | 3D pose estimation · HumanEva-II Combo | mean Euclidean joint position error (mm): 99 -> 79.3 | Helped |
| Use monocular rectified spatiotemporal volumes rather than the strongest reported stereo baseline | 3D pose estimation · KTH Multiview Football II | PCP (All parts): 69 -> 79 | Helped |
Closest papers by meaning
- Exploiting Temporal Context for 3D Human Pose Estimation in the Wild (CVPR 2019 · similarity 0.91)
- Learning 3D Human Dynamics From Video (CVPR 2019 · similarity 0.90)
- Recurrent 3D Pose Sequence Machines (CVPR 2017 · similarity 0.89)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.89)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.89)
- Sim2real transfer learning for 3D human pose estimation: motion to the rescue (NeurIPS 2019 · 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.