On the Use of a Low-Cost Embedded System for Face Detection and Recognition

This paper explores the feasibility of using commercially available off-the-shelf components to implement a low-cost embedded system as the core of a facial detection and recognition system. The system is composed of a Raspberry Pi camera module and a Raspberry Pi B+ enhanced by an Intel Neural Compute Stick 2. Four supervised learning models were implemented on the embedded system for face recognition under different conditions to determine the limitations and capabilities of the system, and the best operational conditions. Best results were achieved when using a Multilayer Perceptron (MLP) algorithm and the distance of the subject to the camera was between 0.3 to 1 meters, the illumination factor in the range from 115 to 130 lux and the horizontal face rotation between -5° to +5°.