Python MXNet pix2pix network. Mainly for colorization of black and white images.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
network
report
test
test_img
util
video_examples
.gitignore
README.md
Visualizing_gluon_network.ipynb
__init__.py
cv2_utils.py
main.py
module_calls_diagram.png
options.py
pix2pix_mxnet_image_colorization_1.ipynb
requirements.txt
trainer.py
u_net_generator.jpeg

README.md

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: Video colorization, captured and colorized in real-time using notebook and usb web-cam