jayadityadev/BrainTumorClassification ? reverse-engineered prompt

Build me a brain tumor classification app that can take an MRI scan and tell me whether it looks like Glioma, Meningioma, or Pituitary.

I want the full workflow in place, from image cleanup and preparation, to model training, to running predictions with a simple web page where I can upload a scan and see the result. It should also show a heatmap or highlight area so I can understand what part of the image influenced the prediction.

Use the data preprocessing steps that improve the image first, because the model should work best that way. If there are existing datasets or setup steps needed, wire those in so I can follow a straightforward run process. Include a way to train the model, save the best model, and then use it for inference in the app.

If you need to check current docs or best practices for anything, go ahead and look them up online.