Deep learning for hand parts segmentation

Introduction

This project aims to segment different parts of hands (recognizing different fingers, the palm and wrist).

It uses a UNet trained on artificial hand data generated by LibHand. Its aim is to generalize to images of real hands. Data are provided in 4 channels: RGB+depth.

TODO