University of Technology Sydney - 42028

Deep Learning and Convolutional Neural Networks

Assignment 3

Group 19 - Face the Facts

Contributors

Name	Email	Github	ID
Harrison Cole	email	Hc747	12962712
Jihee Lee	email	GRORY	13826920
Jose Meza	email	jose-meza-garcia	13093099

About

The wearing of face masks in public spaces has become a normal, if not mandatory, aspect of daily life in most countries around the world. The proper usage of face masks has been demonstrated to be an effective measure in controlling the spread of the coronavirus disease (COVID-19).

Our project delivers a tool that is capable of locating faces within images and videos, detecting if face masks are present, and detecting if they are being worn correctly.

This application is an example of a tool that could be deployed in public spaces (i.e., shopping centres, airports, public transport) in order to provide interested parties with analytical and/or monitoring capabilities. This tool does not provide interested parties with the ability to correlate faces with identities.

Dataset

Images were sourced from the following datasets. Each image was resized to 224x224 and converted to JPEG format.

Cabani	Ashish	Flickr	Ours

classes	masked	unmasked	uncovered_nose_and_mouth	uncovered_nose	uncovered_chin
samples	9,688	9,688	4,834	4,834	4,834

Libraries

TensorFlow	MediaPipe	Dlib	OpenCV	scikit-image	Pillow	NumPy	Tkinter

Architecture

Two models were developed using the following modified VGG16 CNN architecture.

Getting Started

Install

1. Clone the repository.
2. Download the models and unzip them in the same directory as the code.
3. Install the required libraries.

pip3 install -r requirements.txt

Run

python3 main.py

Screenshots