Overview / Usage

Face Detection and Identification has become a very vital task in the modern era of social security and access points. Using deep learning has shown a significant improvement in terms of accuracy and reduction in false identification. However, scalability is an issue with deep learning computation hardware. For this end, Intel has introduced neural compute stick. A device that can be plugged and used as an alternative for massive GPU and CPUs called Neural Compute Stick (NCS). In this project, I would like to test this tech and validate if it can be used for live face detection and identification. Also, I would like to provide some insights into the community by writing a blog and making a video tutorial on YouTube about this work.
The goal of this project is to build an efficient, yet scalable face identification using the combination of edge computing and deep learning.

Methodology / Approach

Using Neural Compute Stick, we train an unsupervised way of recognition and later using the same method try and detect the faces as well. The other method is to first identify the faces using simple Haar features and then recognize using deep siamese networks. A more detailed method will be posted once the project is finalized and fabricated...

Technologies Used

Keras
NCS
Raspberry Pi
Camera module for Raspberry Pi