Project: Simple Face-Mask Detector

A Simple Face-Mask Detector built with OpenCV, Keras(Tensorflow) in order to detect face masks in both images and video streams.

TechStack/Framework

OpenCV
Keras(Tensorflow)
MobileNetV2
- Optimizer: Adam
- Loss: Binary Crossentropy
- Metrics: Accuracy
Caffe-based Face Detector

Dataset

The dataset was created by Prajna Bhandary.

The dataset consists of 1,376 images belongs to two classes:

with_mask: 690 images
without_mask: 686 images

The images used were real images of human faces, however artificial face masks were added to create an artificial (but still real-world applicable) dataset.

Prerequisites

All the dependencies and required libraries are included in the file requirements.txt.

Note: packages.txt is importing streamlit-needed exnternal dependencies. Ignore it if you are using Ananconda venv.

Results

The model gave 98% accuracy for Face Mask Detection after training.
See also in the following accuracy/loss training curve plot:

Working

To train the Face-Mask Detector, execute the following sample command:

1	python train_detector.py -d Dataset

To detect face masks in an single image, execute the following sample command:

1	python predict_image.py -i .\Examples\Images\image-01.jpg

To detect face masks in video streams, execute the following sample command:

1	python predict_video.py

To find a Web-App version of the Simple Face-Mask Detector, execute the following sample command:

1	streamlit run Simple_Face-Mask_Detector_App.py

1	streamlit run https://raw.githubusercontent.com/Ch-i-Yu/Simple-Face-Mask-Detector/main/Simple_Face-Mask_Detector_App.py

or simply visit Streamlit App Link.

Inspiration

This simplified, hello-world like project for a greenhand in Computer vision is inspired by the efforts of Adrian Rosebrock and Prajna Bhandary.

Special Thanks to...

Thanks for my awesome guys, Xiyu L. and Feng J. who helped me to test my tiny project and record the lovely gif in the front of this page.