babajidealao-knn/eye-disease-classification ? reverse-engineered prompt

Build me a simple eye disease classification notebook that trains a deep learning model on retinal images and predicts one of four classes, cataract, glaucoma, normal, or diabetic retinopathy.

I want it to load the dataset, resize images to 224 by 224, split into training and validation, and train a basic CNN that can classify a new eye image from a folder or uploaded file. Please include the full workflow in the notebook, data loading, preprocessing, model training, evaluation, and a few example predictions with clear output so I can see how well it works.

Keep it straightforward and easy to run in Jupyter. If you need to check current best practices for the libraries or notebook setup, look up the latest docs online.