Face-Mask-Detection-Python

Face Mask Detector Using Python3 OpenCV and Tensorflow

Dataset

The dataset was made by Prajna Bhandary The dataset consists of 1,376 images

• with_mask : 690 images
• without_mask : 686 images

Libraries Used

• Tensorflow
• Keras
• Sklearn's train_test_split
• os
• imutils
• numpy
• MobileNetV2
• OpenCV
• Caffe Model OpenCV's face detector

Files

• dataset - Consists of the dataset we will train the model on
• caffe - Consists of the caffe face detection model
• generating_dataset.py - Converts data into labels and numpy arrays and pre-process it for mobilenetv2
• training_model.py - Trains the model with output from generating_dataset.py file
• detecting_mask.py - Uses the Caffe face dectection model to detect faces and pass it to the trained model for predictions
• model - Trained Model saved from training_model.py file

Running

• Run generating_dataset.py to generate data and labels from dataset, it will convert images into array and preprocess it for the mobilenetv2 model save the data and labels in the same directory with names data.npy & labels.npy
• Run training_model.py to load the saved data.npy & labels.npy files and the data is augmenented and uses the base model MobileNetV2 and top of model are few layers of Pooling, Flatten, Dense and 50% Dropout, the model is compiled using Adam optimizer with 32 Batch Size and the model is saved as model file in the same directory
• Run detecting_mask.py file to load the model file and load the Caffe Face Detection model and use it to detect face using webcam stream and using the cropped part from face as input to our trained model.

Accuracy

• Training Accuracy - 99.81%
• Validation Accuracy - 99.64%

Acknowledments

• Prajna Bhandary for the dataset
• PyImageSearch for the caffe model tutorial