ColorNet
lastmod 2026-11-10created 2019-02-122 edits
ColorNet is a neural network implementation that adds color to black-and-white (grayscale) photos. It implements in TensorFlow the approach of the Automatic Colorization article, which uses intermediate features of a pretrained VGG classification network to predict the color components of each pixel; it keeps the luminance channel as-is, estimates only the color channels, and then combines them. It is one of the early public implementations of deep learning-based automatic colorization.
Neural Network to colorize grayscale images.