This is a simple CNN used to classify the PneumoniaMNIST dataset consisting of chest X-rays.
My goal was to be able to perform a binary classification of if a chest X-ray probably has Pneumonia. I tried to rewrite the code to be modular and easily scalable.
This achieved 81% test accuracy and 0.93 ROC-AUC over over 5 epochs (final BCE loss = 0.11).
Some comments are a bit messy, just included to help remind me certain functions.