In the current scenario with COVID-19 virus impacting every country, we need a reliable check for estimation that a person is affected by the virus or not. While our hack is not exactly a reliable or go-to estimate (because it's prediction, after all, and the chances of false-positive / negative are high), it still aids doctors and medical professionals a great deal in taking chest X-Ray scans or CT scans and guessing if the person is affected by the COVID-19 virus or not.
The end to end integration with image upload to model prediction put us in a pinch. The keras load_model() function wasn't loading our weights and preparing JS functions for updating image cards after image upload was a bit tricky too. While the latter was handled using a simple DOM manipulation, the former required us to store several trained models which took a lot of time in training