Whiten Color Transfer in Tensorflow

Objective

This is an implementation of Whiten Colour Transfer in Tensorflow. A computer program that can paint over any image with the style of any other image instantly! No retraining required! It incorporates multiple algorithms and settings to allow the user to customize​ the artwork generated. These modes and settings are as follows… 1)WCT – Whiten Colour Transfer (Transfers relevant features between images) 2) adaIN – Arbitrary Style Transfer (Transfers lighting, colour, and relevant textures between images) 3)swap5 – Swap5 algorithm (Transfers all features between images) 4) KPT – Keep Original Colours (Transfers the style, but not the colour)

Meeting

Monday 9-4 pm

Students

Mentor(s)

Blog

Week 1

May, 20 2019


What can it do so far?

Stylise images and videos using a Command Line Interface (command prompt with python)

Stylise images using a Graphical User Interface (GUI python)

Train standard (not optimised) models in Python

What Happened This week?

Retrained models in python unsuccessfully. Try it again later.

To do:

  1. Retrain models in python
  2. Compress and distil models in python
  3. Convert Tensorflow python code to Tensorflow.js code
  4. Make a javascript implementation of Koby_WCT-TF and make it run locally. (on your PC)
  5. Add that to wordpress via wp-code. (see plugins wp-code)
  6. Make it pretty
  7. Make it mobile via web
  8. Make it a native App for android
  9. Make it a native App for iOS

Week 2

May, 30 2019


What can it do so far?

Stylise images and videos using a Command Line Interface (command prompt with python)

Stylise images using a Graphical User Interface (GUI python)

Train standard (not optimised) models in Python

To do:

  1. Compress and distil models in python
  2. Convert Tensorflow python code to Tensorflow.js code
  3. Make a javascript implementation of Koby_WCT-TF and make it run locally. (on your PC)
  4. Add that to wordpress via wp-code. (see plugins wp-code)
  5. Make it pretty
  6. Make it mobile via web
  7. Make it a native App for android
  8. Make it a native App for iOS