Research map / 3D vision / Human pose estimation
Occlusion-aware Hand Pose Estimation Using Hierarchical Mixture Density Network
Qi Ye, Tae-Kyun Kim
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 visibility hierarchy (HMDN instead of MDN), J=20 | MDN · EgoBigHand | Visible error (mm): 34.0 -> 30.7 | Improved |
| add visibility hierarchy (HMDN instead of MDN), J=20 | MDN · EgoBigHand | Occluded error (mm): 36.4 -> 34.4 | Improved |
| HMDN instead of SGN, occluded joints | SGN · EgoBigHand | Occluded error (mm): 36.5 -> 34.4 | Improved |
| HMDN instead of SGN, visible joints | SGN · EgoBigHand | Visible error (mm): 32.8 -> 30.5 | Improved |
| mixture density (MDN) instead of single Gaussian | SGN · EgoBigHand | Occluded min-distance error (mm): 38.9 -> 34.8 | Improved |
| HMDN instead of MDN on min-distance error to multiple labels | MDN · EgoBigHand (J=10) | Occluded min-distance error (mm): 34.8 -> 34.6 | No clear effect |
| HMDN instead of MDN on min-distance error, J=30 | MDN · EgoBigHand (J=30) | Occluded min-distance error (mm): 35.0 -> 34.5 | Improved |
| increase number of Gaussians from 10 to 30 (ablation) | HMDN · EgoBigHand | Occluded error (mm): 34.8 -> 34.2 | No clear effect |
| sample from visibility distribution in training (HMDN_soft vs hard), J=10 (ablation) | HMDN_hard · EgoBigHand | Visible error (mm): 32.2 -> 30.5 | Improved |
| sample from visibility distribution in training (soft), J=10 (ablation) | HMDN_hard · EgoBigHand | Occluded error (mm): 35.8 -> 34.8 | Improved |
| sample from visibility distribution in training (soft), J=20 (ablation) | HMDN_hard · EgoBigHand | Occluded error (mm): 35.9 -> 34.4 | Improved |
| sample from visibility distribution in training (soft), J=30 (ablation) | HMDN_hard · EgoBigHand | Occluded error (mm): 36.4 -> 34.2 | Improved |
Closest papers by meaning
- Harmonious Feature Learning for Interactive Hand-Object Pose Estimation (CVPR 2023 · similarity 0.89)
- Depth-Based 3D Hand Pose Estimation: From Current Achievements to Future Goals (CVPR 2018 · similarity 0.89)
- Self-Supervised 3D Hand Pose Estimation Through Training by Fitting (CVPR 2019 · similarity 0.88)
- DeepHand: Robust Hand Pose Estimation by Completing a Matrix Imputed With Deep Features (CVPR 2016 · similarity 0.88)
- Robust Hand Pose Estimation During the Interaction With an Unknown Object (ICCV 2017 · similarity 0.88)
- HandOccNet: Occlusion-Robust 3D Hand Mesh Estimation Network (CVPR 2022 · 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.
