Cascade face classification framework based mask detection of real time images using deep learning approach

Detecting the faces with mask in efficient way is becoming incredibly valuable, as it can be used in COVID-19 situation as well as it can play a major role to follow and identify criminals or terrorists. Because of the high occlusions that result in the loss of face information, masked face identification is a far more difficult process. The proposed scheme employed a way to states that a person had worn a mass or not. A simple convolutional neural network (CNN) model is developed using deep learning and combining Tensor Flow, Keras, Scikit-learn, and Open-CV to make the system as accurate as possible. Moreover, using the Flask framework and a webcam, a real-time detection application has been created that is capable of accurately predicting a person has wear a mask and not, as well as showing a label in the right corner of the image. Categorical cross-entropy is applied as a loss function. The proposed model’s validation accuracy is 97.45 percent. Anyone on the live stream who is not wearing a mask has a rectangle around their face that is red, with the dialogue "NO MASK," while anyone who is wearing a mask gets a rectangle around their face that is green.