This group project is meant to simulate a real-world workplace situation. Our team (of three) set out to predict COVID hotspots based on analyzing Twitter photos showing people wearing face masks, then predict infection rates.

Our model uses a neural network algorithm trained on data from Kaggle to classify Twitter photos that contain mask-wearers. We utilized Tweepy, MobilenetV2, BeautifulSoup, OpenCV, and Keras. The model’s precision improved from 0.55 to 0.77.

Next steps are to integrate the Twitter image classifier with GIS to map the results, which would enable public health agencies to prepare for outbreaks and hopefully slow the spread.

Please follow the progress here.