a master’s thesis in the College of Engineering at the University of Basrah discussing (Recognizing multiple faces in real time using deep learning)

The thesis presented by the student Nour Faleh Abdel Hassan dealt with the use of YOL0 frames for real-time face recognition instead of using them to detect objects on an embedded device The frameworks are integrated into an embedded GPU system. The proposed system includes two distinct implementations of the YOLO framework. First, YOLOv4tiny is used, which can detect two faces in real time with less FPS.

Secondly, YOLOv5s are designed using Docker and DeepStream and real-time medical mask detection for up to thirteen faces at a high frame rate based on experimental data.

The thesis aims to develop a monitoring system capable of recognizing multiple faces and discovering multiple masks. And the medical mask detection system in addition to a temperature sensor and an alarm device that is activated in case of sensitivity to high human heat.

The thesis concluded that the proposed models for face recognition and mask detection and inference were trained using CUDA on Google Colab, MacBook and Jetson Nano Developer kit.

The MobileNetv2 classifier is trained using the deep learning libraries TensorFlow and Keras and OpenCV.