Detection Face with Mask and Without Mask Using Open CV

Abstract

Facial mask detection is a relevant application of computer vision, the use of which has become crucial in the context of preserving safety and health measures of the population. The classical deep learning models like Convolutional Neural Networks (CNNs) are good at local features extraction and are weak in capturing long-ranged image dependencies. To overcome such weaknesses, this article will suggest one hybrid deep learning network that incorporates EfficientNet to extract features efficiently and Vision Transformer (VIT) to process features in global contexts with the help of self-attention mechanisms. The suggested system is trained and tested on the publicly available set of 7,553 images of masked and unmasked faces. Extensive experimental research has shown that the hybrid EfficientNet VIT model is more accurate, precise, recalls and F1-score than the traditional ANN, CNN and FNN models. The approach to be proposed has a greater strength in changing lighting conditions, background complexity, and face orientations. The model can be used in real-time execution on the surveillance systems, schools, airports, and in healthcare facilities.