A basic convolutional neural network (CNN) approach to image colorization.
Keras, TensorFlow, Scikit-Learn.
Colorization refers to the task of taking a grayscale image as input and producing a "realistically" colored version of the input image as output.
The implemented model is roughly based off of the model described here. Essentially, it consists of a series of convolutional layers followed by alternating upsampling and convolutional layers.
The following will generate colorized versions of the provided image dataset:
python deep_colorization.py