The notebook requires a kaggle environment with the following 3 datasets added in the /inputs/ directory
Just create a new Kaggle notebook and click on Add Data in the top right corner. Also set the accelerator to be the GPU. Training becomes much faster that way.
From here onwards, just run all the cells using a GPU accelerator and wait for 5-6 minutes for the model to be trained and evaluated.
The saved model could be deployed onto a webpage and/or IoT hardware. Since it performs face detection AND mask detection, it can perform as a "security camera" to detect COVID violations. Dystopian I know...
References
- This notebook by Nagesh Singh
- This excellent guide on transfer learning
- Keras Docs
- Haarcascades