Research map / 3D vision / Human pose estimation
Hand Image Understanding via Deep Multi-Task Learning
Xiong Zhang, Hongsheng Huang, Jianchao Tan, Hongmin Xu, Cheng Yang, Guozhu Peng, Lei Wang, Ji Liu
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 |
|---|---|---|---|
| use 4 cascaded MTL stacks instead of 1 (ablation) | HIU-DMTL · HIU-Data 2D pose | 2D Pose score: 0.837 -> 0.866 | Improved |
| use 4 cascaded MTL stacks instead of 1 (ablation) | HIU-DMTL · HIU-Data hand mask | Hand Mask score: 0.948 -> 0.974 | Improved |
| use 4 cascaded stacks instead of 1 (ablation) | HIU-DMTL · FreiHAND hand mesh | Hand Mesh score: 0.842 -> 0.856 | Improved |
| use 4 stacks instead of 1 (ablation) | HIU-DMTL · FreiHAND 3D pose | 3D Pose score: 0.852 -> 0.860 | Improved |
| go beyond 4 stacks (8 stacks) (ablation) | HIU-DMTL · HIU-Data 2D pose | 2D Pose score: 0.867 -> 0.870 | No clear effect |
| train 2D pose, mask and mesh jointly instead of mask only (ablation) | HIU-DMTL · HIU-Data/FreiHAND hand mask | Hand Mask score: 0.863 -> 0.907 | Improved |
| train 2D pose, mask and mesh jointly instead of mesh only (ablation) | HIU-DMTL · FreiHAND 3D pose | 3D Pose score: 0.802 -> 0.821 | Improved |
| train all three tasks instead of 2D pose only (ablation) | HIU-DMTL · FreiHAND 2D pose | 2D Pose score: 0.752 -> 0.802 | Improved |
| add task attention module (TAM) in 4-stack model (ablation) | HIU-DMTL · HIU-Data 2D pose | 2D Pose score: 0.793 -> 0.866 | Improved |
| add TAM in 4-stack model (ablation) | HIU-DMTL · FreiHAND hand mesh | Hand Mesh score: 0.808 -> 0.856 | Improved |
| add TAM in 1-stack model (ablation) | HIU-DMTL · HIU-Data 2D pose | 2D Pose score: 0.821 -> 0.837 | Improved |
| add TAM in 1-stack model (ablation) | HIU-DMTL · FreiHAND hand mesh | Hand Mesh score: 0.826 -> 0.842 | Improved |
Closest papers by meaning
- 3D Hand Shape and Pose Estimation From a Single RGB Image (CVPR 2019 · similarity 0.89)
- Model-Based 3D Hand Reconstruction via Self-Supervised Learning (CVPR 2021 · similarity 0.88)
- Robust 3D Hand Pose Estimation in Single Depth Images: From Single-View CNN to Multi-View CNNs (CVPR 2016 · similarity 0.88)
- HandVoxNet: Deep Voxel-Based Network for 3D Hand Shape and Pose Estimation From a Single Depth Map (CVPR 2020 · similarity 0.88)
- Collaborative Learning for Hand and Object Reconstruction With Attention-Guided Graph Convolution (CVPR 2022 · similarity 0.88)
- Reconstructing Hands in 3D with Transformers (CVPR 2024 · similarity 0.86)
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.
