This project is a replica in Python of this article :
https://handtracker.mpi-inf.mpg.de/projects/GANeratedHands/content/GANeratedHands_CVPR2018.pdf
This project doesn't have similar real-time performances as the original one made with Caffe and C++. Thus, it is for educational purpose only.
This project also contains files from this project :
https://github.com/Ninebell/GaneratedHandsForReal_TIME
Files reused for this project : projLayer.py & regnet.py. Both have been modified significantly.
I encourage you to visit both projects and support their creators.
The purpose of this project is to replicate the architecture of the Residual Network found in the article. The ResNet (or RegNet) is a Deep Convolutionnal Neural Network designed to track 21 points from an image of a hand.