An Evaluation of MTCNN in Face Recognition Algorithms for Effective Detection in Masked Scenarios in Real Time Video Surveillance

Facial expression analysis has been more widely accepted and used in a variety of domains, including computer vision, psychology, neuroinformatics, and artificial intelligence. However, Face identification poses challenges due to occlusions and variations in facial alignments., like the one that resulted from wearing PPE during the COVID-19 pandemic. Since masks cast a shadow in the areas where the primary means of positioning and identifying obstacles for model construction are located, the analysis of faces wearing masks has an impact on various facial recognition models. The experiments in facial recognition analysis using Multitask Convolutional Neural Networks (MTCNN) obscured by mask utilization is presented in this research. 1173 images were utilized in the MTCNN's training process. According to the experimental results, a Multitask Convolutional Neural Network trained with face images under partial occlusion in this case, a mask allows for up to 99% facial recognition detection; in particular, the output function's detection of the eye area performs exceptionally well in both mask-free and mask-on scenarios in real time video survilleance.