EmirMuhammetARAN/Brain-Tumor-MRI-Classification---99-ACC---Grad-CAM- — reverse-engineered prompt

Build me a simple medical image AI demo for brain tumor MRI classification. I want a user to upload a JPG or PNG MRI scan, then the app should predict one of four results, glioma, meningioma, no tumor, or pituitary, with confidence scores that are easy to understand.

Please also show a Grad CAM heatmap over the image so people can see what area the model focused on. Add a short interpretation guide, like high confidence, medium confidence, and low confidence where expert review is needed. Make it clear this is for research and education, not a replacement for a doctor.

Use the existing trained weights if available, and include a clean web interface plus an API endpoint for prediction. I’d also like basic evaluation pages or scripts that show accuracy, confusion matrix, precision, recall, F1 score, and clinical style metrics like sensitivity and specificity.

Make it runnable locally, and also ready to deploy with Docker or on Hugging Face Spaces. Look up current docs online if needed.