Title: Face Mask Wearing Manner Detection

Due to the COVID-19 pandemic effect, masks have become a successful weapon to defend ourselves against the spread of viruses. Most people wear their masks in inappropriate ways,making them more vulnerable to the spread of viruses. This project aims to predict whether a person is appropriately wearing a face mask.

A Caffe model based on the Single Shot-Multibox Detector (SSD) uses ResNet-10 as the backbone is used with OpenCV to detect the faces in an image. The ResNet50 pre-trained transfer learning model is used to train the Tensorflow model over the dataset consisting of more than 5000 images of eight classes. The eight classes which will be predicted are ‘Correctly wearing the Mask’, ’No Mask’, ‘Mask on the forehead, ‘Mask under the chin’, ‘Mask above the chin’, ‘Mask on the tip of the nose’, ‘Mask under the nose’ and ‘Mask hanging from one ear. The model is trained to get an accuracy of around 80%, which works well with the validation data. The models trained to load, and the predictions are applied onthe images fetched from files or from the webcam which OpenCV preprocesses. The face which the Caffe model detects is indicated by the rectangular bounding boxes drawn in the picture. The preprocessed portion of the image using the Caffe model and the trained ResNet50 model is applied to that portion of the image. It makes predictions on that image (The confidence of the forecast is also displayed with the rectangular bounding box over the face). The colour of the bounding box will be drawn green if the mask has been appropriately worn, or else the box’s colour would remain red. The app is deployed using the streamlet 0.82.0 framework. The deployed app is capable of predicting the classes on the uploaded images and also capable of predicting the types through a webcam.

Team Members

Mr. C. Rajeswari
Assistant Professor, Department of Computer Science and Engineering
Team Leader
Mr.R.Aswin Balaji
Team Leader
Mr.Gelli Chakradhar
Team Member
Mr.S.Vinu Balan
I Year CSE
Team Member


Admission Open 2022