Pix2pix for image colorization

Python 3.6 and MXNet 1.1

This is an attempt to implement Pix2pix network using Python and MXNet Mainly used to colorize black and white images, specifically videos.

The network is configured right now to work with input of 1x3x256x256 images. An image will be split into lightness and a,b channels, the lightness channel will serve as an input to generator which will attempt to produce a 1x2x256x256 a,b channels tensor that we will feed into the discriminator in a sequence with original a,b channels.

Metrics visualization

tensorboard --logdir=./logs/train

Code is based on

https://github.com/affinelayer/pix2pix-tensorflow https://github.com/zackchase/mxnet-the-straight-dope/blob/master/chapter14_generative-adversarial-networks/pixel2pixel.ipynb

pix2pix paper

https://arxiv.org/abs/1611.07004

#

Video colorization example: