Grayscale images colorization with convolutional neural networks

Previous approaches to the colorization of grayscale images rely on human manual labor and often produce desaturated results that are not likely to be true colorizations. Inspired by Matías Richart’s paper, we proposed an automatic approach based on deep neural networks to color the image in grayscale. We have studied several models, approaches and loss functions to understand the best practices for producing a plausible colorization. By noting that some loss functions work better than others, we used the VGG-16 CNN model based on the classification with the loss of cross-entropy. The experiment shows that our model can produce a plausible colorization.