Mask Detection with an Improved Faster R-CNN

The Coronavirus pandemic has underscored the critical need for effective mask detection to mitigate the virus's spread. Despite global vaccination efforts, the virus persists, necessitating continued mask usage, especially in public spaces. This study proposes an enhanced faster Region-based convolutional Neural Network (RCNN) model incorporating ResNet-50 and Feature Pyramid (FPN) as backbone networks. The model leverages a RPN to efficiently identify mask-wearing statuses. A new dataset, combining images from MaskedFace-Net and public sources was used to train and evaluate the model, which include three categories; mask, no mask and false mask (mask worn wrongly). The proposed model achieved superior performance with a mean average precision (mAP) of 98.4%, surpassing traditional backbones. The ResNet-50 with the FPN configuration demonstrated balanced accuracy and robust feature extraction capabilities, significantly outperforming other models in various metrics. The enhanced integration of the model provides a highly effective solution for mask detection, ensuring reliable performance in diverse real-world scenarios.