Face detection and recognition has been one of the most complex AI tasks to achieve. There have been quite a few algorithms to achieve this, yet there remains a great scope for exploration. This research offers collaboration of two state-of-the-art algorithms - YOLO & FaceNet with a classiffer to detect and recognize faces in images and real-time live streaming from webcam. It proposes a 3-stage architecture built and implemented from scratch and takes an effort to create a custom data set from scratch.
This project proposes to give a new approach to face
recognition problem. The project is taking up 2 state-of-the-art algorithms namely
YOLO and FaceNet along with a classification algorithm and make a 3 staged
face recognition system. Its also proposed to make this system work on still images as
well as live streaming from webcam. And the data used to train this network is
custom and gathered from scratch.
To build the proposed system, break down the problem statement into 3 major parts:
Face detection, Face Recognition-(Embedding Calculation)
and Face Classification. The data passed will go through these 3 stages
before outputing it to the end user.
Youtube Link: https://www.youtube.com/watch?v=83F3ZW48Ox0&feature=youtu.be