Multistage Framework for Automatic Face Mask Detection Using Deep Learning

Algorithm 1

Face mask detector algorithm.

	Step 1: Train the model for face mask detection
	Step 1.1: Load the dataset[RMFD, IMFD, Kaggle, search engine api]
	Step 1.2: Extract Class labels.[A : Mask b: Without Mask]
	Step 1.3: Preprocess the input data.
	Step 1.4: Label the data.
	Step 1.5: Split the dataset suitably, for training and testing.
	Step 1.6: Load pre-trained model as the base model of backbone.
	Step 1.7: Add the Head of backbone model by adding cnn layers above the base model.
	Step 1.8: Freeze the summed model to avoid updation of weights.
	Step 1.9: Train model for suitable epochs based on accuracy.
	Step 1.10: Save the model for future usage.
	Step 2: Load the image/recorded video/live video feed
	Step 3: Locate faces with the help of a CNN based face detection[Haar, MTCNN and DNN algorithm based on input type]
	Step 4: Recognise Region of Interest (RoI) in faces by applying segmentation algorithm.
	Step 5: Input it to the saved CNN model to classify faces as face mask or No-mask.
	Step 6: Verify faces classified as ’face-mask’ have the face-mask worn properly or not with suitable alert message
	Step 7: End