Download/Clone the repository, navigate to the folder and:
Run the Python notebook, to generate the model.json file, and the model_weights.h5 file.
Then run command:
python main.py
Serves on https://localhost:5000.
Change video/image stream for face recognition from the file camera.py. Navigate to the line:
self.video = cv2.VideoCapture("presidential_debate.mp4")
Replace presidential_debate.mp4 with a file of your choice, or enter 0 for webcam input.
You can get it off Kaggle, here.
I used the distribution available on Coursera, where they already divide the dataset according to its class into its specific folder.