Photoshop Fast Style Transfer Using Machine Learning 🎨

A user-friendly tool to use fast-style-transfer for stylizing images directly inside Photoshop.

Requirements

Based on the fast style transfer repository the machine learning solution requires tensorflow, Pillow, numpy and scipy to be installed on your python environment.

• Therefore there is "requirements.txt" that includes all the required libraries and can be installed simply by running this pip command pip install -r requirements.txt
• Make sure to have python.exe and scripts/pip.exe is already in your system PATH environment variable.
• This solution is using tensorflow-gpu which requires by default to have a supporting CUDA GPU and cuDNN installed.
• If you are going to install the same tensorflow-gpu version that is mentioned in the requirement.txt file we recommend to use CUDA v9.0 with the compatible version of cuDNN.
• In case of not having a supporting GPU, you can still use tensorflow the cpu version but the image processing time will be longer.

Installation

• After fulfilling the requirements all what you need is just to copy the Presets folder inside your photoshop root directory and accept the overwriting.
• Download the available models provided by the fast-style-transfer author from here https://drive.google.com/drive/folders/0B9jhaT37ydSyRk9UX0wwX3BpMzQ
• copy the downloaded models .ckpt files inside the models folder as following "Presets/Scripts/StyleTransfer/models/".
• Restart your photoshop if you have a running session already.
• Run the script from File > Scripts > [3Deep] StyleTransfer.
• Choose your favorite style from the avaliable styles.
• To expand your collection of styles please check the fast-style-transfer repoistory for how to train a new model instructions.

Links

• cuDNN | https://developer.nvidia.com/cudnn
• CUDA | https://developer.nvidia.com/cuda-toolkit-archive

Authors

• Photoshop tool (.jsx and .py) 3deep.org | author Mahmoud Hesham.
• Fast Style Transfer Repository "Fast Style Transfer in TensorFlow" | author lengstrom.

License

• All the scripts/tools/software-bridges developed by 3Deep are free for academic and non-commercial use only.

The ML-solution license stated from the fast-style-transfer author:

Copyright (c) 2016 Logan Engstrom. Contact me for commercial use (or rather any use that is not academic research) (email: engstrom at my university's domain dot edu). Free for research use, as long as proper attribution is given and this copyright notice is retained.