The Project was implemented using a dataset consisting of 4120 images collected from Kaggle,Self captured images, Celebrity Face Dataset and using Bing Search API. The Project uses a convolutional neural network to classify people with and without masks using 3 different pre trained models used for transfer learning namely VGG-16, VGG-19 and MobileNetv2 with the last one giving the best results, having a training accuracy of 99.78% and a test accuracy of 97.47%. The images were preprocessed, converted to arrays, normalized and data augmentation was done as there was a little data for the model to train. The model was trained and saved using Keras, Tensorflow and implemented on real-time and static images and videos using OpenCV. The model was then deployed locally using Flask in the backend and HTML for the Frontend.
gshivansh2001/Face-Mask-Detection
Using machine learning to predict if a person is wearing a face mask or not. The application can be used through this link: https://huggingface.co/spaces/shivanshg/Face_Mask_Detection
Jupyter Notebook