vitorrlopes / mask-detector

During Covid19 quarantine i decided to implement my own version of face mask detector using Deep Learning models just for fun.

mask-detector

During Covid19 quarantine i decided to implement my own version of face mask detector using OpenCV, Keras and Tensorflow Deep Learning models just for fun.

Checkout the code:

git clone https://github.com/fnandocontreras/mask-detector.git

Install Requirements

To be able to run the code, before you need to install the required packages:

To install using anaconda run the following commands:

conda env create -f environment.yml
conda activate maskdetector

The file "environment.yml" contains the required packages for running mask-detector and after running the commands above, anaconda will create a virtual environment name "maskdetector" with all the requirements specified in the env file. If you don't have anaconda you can install the required packages by hand using PIP:

pip install numpy tensorflow keras mtcnn pillow opencv

After installation you are ready to run the mask detector program

Running the mask detector

For running the mask detector using your webcam live streaming run the following command:

python app.py

For running the mask detector on a video file run the following command:

python app.py --source <your-video-filename>

For some videos made with a smartphone you can experience that the image is rotated 90 degrees, this is because the mask detector is not checking orientation. To fix this you can pass the argument: "rotate" like:

python app.py --source <your-video-filename> --rotate

Recording

You can also record the video streaming by passing the argument "--rec" like:

python app.py --rec

It will record the video streaming in a video file named: "record.avi"

Training the classification model

If you are curious to see how i trained the model using tensorflow and keras you can see the notebook: training_model.ipynb