Research map / 3D vision / Human pose estimation
Learning Hand Articulations by Hallucinating Heat Distribution
Chiho Choi, Sangpil Kim, Karthik Ramani
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 multi-task per-finger branches instead of a single holistic network (ablation) | Depth network · Synthetic hand dataset | mean joint angle error | Improved |
| group fingers anatomically (T-IMR-L) instead of dividing into six sub-tasks (ablation) | Depth network · Synthetic hand dataset | mean joint angle error | Improved |
| weight thumb loss 3 times larger (alpha = 3) (ablation) | Depth network · Synthetic hand dataset | mean joint angle error | Improved |
| add hallucinated heat distribution features and refinement network (ablation) | Depth-alone baseline · Synthetic hand dataset | mean joint angle error | Improved |
| add hallucinated heat distribution features and refinement network (ablation) | Depth-alone baseline · NYU | fraction of frames within distance threshold | Improved |
| use hallucinated features instead of ideal heat distribution network features (ablation) | Refinement network · Synthetic hand dataset | mean joint angle error | No clear effect |
| concatenate depth features after conv6 and heat features after conv5 versus other layer combinations (ablation) | Refinement network · Synthetic hand dataset | mean angle error | Improved |
| use unpooling versus deconvolution for segmentation upsampling (ablation) | Localization network · HandNet | segmentation accuracy and time | No clear effect |
| use different set of heat source vertices J (ablation) | Heat distribution descriptor · Synthetic hand dataset | regression accuracy | No clear effect |
| proposed method versus generative method [23] | Ours · MSRA14 | mean distance error (mm): 9.2 -> 20.0 | Hurt |
| proposed method versus discriminative method [7] | Ours · MSRA14 | mean distance error (mm): 22.7 -> 20.0 | Improved |
Closest papers by meaning
- Robust 3D Hand Pose Estimation in Single Depth Images: From Single-View CNN to Multi-View CNNs (CVPR 2016 · similarity 0.90)
- DeepHand: Robust Hand Pose Estimation by Completing a Matrix Imputed With Deep Features (CVPR 2016 · similarity 0.89)
- Robust Hand Pose Estimation During the Interaction With an Unknown Object (ICCV 2017 · similarity 0.89)
- Dense 3D Regression for Hand Pose Estimation (CVPR 2018 · similarity 0.88)
- Learning to Estimate 3D Hand Pose From Single RGB Images (ICCV 2017 · similarity 0.88)
- Aligning Latent Spaces for 3D Hand Pose Estimation (ICCV 2019 · 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.
