This project uses an autoencoder model to colorize grayscale images by predicting the color channels from the lightness channel in the Lab color space.
The model takes a grayscale image (L channel) as input and outputs the two color channels (a and b), which are combined to reconstruct a color image.
You can train the model, visualize results, and evaluate on real grayscale images using the provided scripts.