Silvano315/Pneumonia_Detection ? reverse-engineered prompt

Build me a Jupyter notebook project that detects pneumonia from chest X ray images using a binary classification approach.

I want the workflow to cover basic image exploration and preprocessing, then train one CNN from scratch and compare it with a transfer learning version. It should handle the train, validation, and test folders with Normal and Pneumonia classes, deal with the class imbalance, and show the training results clearly with accuracy and loss plots.

Please also add Grad CAM so I can see which parts of the X ray are influencing the prediction, and include a simple demo app where I can upload an image, get the predicted label, and see the heatmap overlay. Make the whole thing easy to run from the notebooks, and if you need to check current docs or examples online, go ahead.