Medical Face Mask Application Tool

This is a tool that can be used to create masked versions of existing image datasets and study visual recognition problems, such as face identification/verification or emotion recognition in the presence of medical face masks.

It also has fast ML inference & processing accelerated even on common hardware and can be utilized in real-time applications.

Input	Output

Prerequisites

Install the required packages below:

pip install mediapipe
pip install opencv-python
pip install tqdm

How to run

Image

Apply mask on a single image:

python mask_on_img.py IMAGE_NAME

Videos (preserving audio)

Apply mask on multiple videos, by creating a folder 'videos' in the same directory and running:

./mask.sh

Webcam

Apply mask in real-time on multiple faces with webcam input:

python mask_webcam.py

How it works

Original	Face Mesh	Jawlines	Masked

• Face Mesh tracking with Google's MediaPipe Face Mesh.

• Jawline extraction in lines 43-69 in mask.py and face_mesh_model.png