Event Details
Deployment of a Real-Time Face Mask Classification System Using Browser Webcam Streaming and FastAPI
Presenter: Yazhini Venkatraman
Supervisor:
Date: Mon, March 16, 2026
Time: 12:30:00 - 00:00:00
Place: Online via Zoom
ABSTRACT
Meeting link: https://uvic.zoom.us/j/83107655614?pwd=UXxZdNOzlUk5tNluZCjPrfQh4jQAJF.1
Meeting ID: 831 0765 5614
Password: 565656
One tap mobile
+16475580588,,83107655614# Canada
+17789072071,,83107655614# Canada
Dial by your location
+1 647 558 0588 Canada
+1 778 907 2071 Canada
Meeting ID: 831 0765 5614
Find your local number: https://uvic.zoom.us/u/kktbdbawv
Abstract: This project presents a real-time face mask classification system designed to support safety monitoring in environments such as workplaces, institutions, and healthcare facilities. The system detects faces using a Caffe based Single Shot Detector and classifies mask usage into four categories: with mask, without mask, with N95 mask, and improper mask. A curated dataset containing 14,857 images across four classes was prepared and preprocessed through face detection, cropping, resizing to 224×224, normalization, and augmentation techniques such as brightness / contrast variation, geometric transformations and blur to improve robustness under different lighting and orientation conditions.
The classification model is built using MobileNet pretrained on the ImageNet dataset as a feature extractor with a custom classifier head. Training uses the Adam optimizer and early stopping with patience values between 1 - 7 is applied to prevent overfitting. Model performance is evaluated using accuracy, precision, recall, F1-score, and confusion matrix analysis. Experimental results show that the proposed system achieves an overall classification accuracy of approximately 97.55% (patience = 4) on the test dataset. A browser based webcam interface with a FastAPI backend is implemented to demonstrate real-time face detection and mask classification on standard webcam.