/ Machine Learning

Track and extract the optimal face in multi-target faces

Track and extract the optimal face in multi-target faces

Face Detection & Tracking & Extract

This project can detect , track and extract the optimal face in multi-target faces (exclude side face and select the optimal face).

Introduction

  • Dependencies:
    • Python 3.5+
    • Tensorflow
    • MTCNN
    • Dlib
    • Scikit-learn
    • Numpy
    • Scikit-image

Run

  • To run the python version of the code you have to put all the input videos in one folder like /home/admin/videos and then provide the path of that folder as command line argument:
python3 start.py /home/admin/videos 
  • Then you can find faces extracted stored in the floder ./facepics .
  • If you want to draw 5 face landmarks on the face extracted,you can make the argument face_landmarks to be True
python3 start.py /home/admin/videos --face_landmarks True

What can this project do?

  • You can run it to extract the optimal face for everyone from a lot of videos and use it as a training set for CNN Training.
  • You can also send the extracted face to the backend for Face Recognition.

Results

alt text
alt text

GitHub