GDO-GAZE Project

Report

GDO-GAZE is a gaze estimation software using a kinect (hopefully multiple kinects soon) and a face recognition library. It is developped in Python and requires a GPUn for optimal performance.

Installation

Get the GDO-GAZE source code

git clone https://github.com/dsi-icl/gdo-gaze.git

Requirements

This project was developped using Anaconda.

• Python 3.5+ or Python 2.7
• Linux, Windows or macOS
• pytorch (>=1.0)
• Face_alignment library by Adrian Bulat
• Websocket client
• Opencv
• Pykinect2:
  1. pip install comtypes
  2. Install the Kinect for Windows SDK v2
  3. pip install pykinect2

❗ Use pip in the following directory

• Mac: .../anaconda3/lib/python3.7/site-packages/
• Windows: .../anaconda3/lib/sites-packages/

Launching the software

Open a terminal in Eye-Detection/FaceLandmarksKinect/:

python benchmark.py

Contributors

Development of this project was done by Andrianirina Rakotoharisoa under the supervision of Dr Ovidiu Șerban and Florian Guitton.