This repository aims to understand and implement hand pose estimation methods using Pytorch. Currently, we only provide the modified algorithm proposed in
[1] Oberweger, Markus, Paul Wohlhart, and Vincent Lepetit. “Hands Deep in Deep Learning for Hand Pose Estimation.” https://arxiv.org/abs/1502.06807
- NYU 2014 dataset should be put in
data/nyu14/
To impose the prior of the low-dimensionalty of hand pose, the authors in [1] added a bottle neck layer before the last fully connected layer. The effect of dimension of this layer is provided. I expect that the lower dimensionality yields better results. The result, however, shows that the higher dimensionalty yields better performance. I think it is related to model complexity and other parameters. I will check later again.