This project contains my experiment to apply the Style Transfer algorithm in Deep Learning to produce a new "Thai Modern Artwork".
See more details about this work in my medium post (Style Transfer) How to create a "Thai Modern Artwork" without artistic skills
Below are python libraries that are required to run this code using Python versions 3.7.0:
- matplotlib
- numpy
- Pillow
- pytorch 1.0 (Stable) with CUDA 10.0 (See here for the installation instruction)
- torchvision
- seaborn
This project contains two notebooks which walk you through the experiment process where:
- Analysis_Content_Style_Representation.ipynb - This notebook contains the Analysis of the Content Representation and Style Representations of the images that we will be using in this experiment, and I would highly suggest going over this notebook first to get a better understading of the concepts of the Style Transfer algorithm and the setup for the experiment.
- In case there is an error when trying to render this notebook in github, you can download the html file Analysis_Content_Style_Representation.html
- StyleTransfer_Experiment_Results.ipynb - This notebook contains the Experiment results with various parameter settings
- In case there is an error when trying to render this notebook in github, you can download the html file StyleTransfer_Experiment_Results.html
- Some part of the source code in this project was inspired by a hands-on lab for the "Style Transfer" module of the Udacity's Deep Learning Nanodegree Program and must give credit to them.
- Images used in this experiment were downloaded online. Please find where those images were downloaded in ./images/README.md and must give credit to the owner of those images.
This project is MIT licensed.