agamkachhal/Machine-Learning---Hair-Colour-Prediction-using-Convolutional-Neural-Nets.- ? reverse-engineered prompt

Build me a notebook that trains a convolutional neural network to predict a celebrity’s hair color from images. I already have the dataset split into training, validation, and test files, and the labels are 0, 1, 2, 3 for black, blond, brown, and gray hair.

Use the provided image arrays and label files, load everything cleanly, and make the full workflow easy to run from top to bottom in Jupyter. I want the notebook to include data loading, basic preprocessing, the model, training, validation, and a final test evaluation. Please also show a few example predictions so I can see how well it’s doing.

Keep it simple and practical, since the images are small and the dataset has already been trimmed down. If you need to check anything about the CelebA format or best practices, look up current docs online.