Skip to content
🖼 Tensorflow implementation of a fast feed-forward neural style transfer network
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.


Type Name Latest commit message Commit time
Failed to load latest commit information.

Deep Style Transfer

Tensorflow implementation of the fast feed-forward neural style transfer network by Johnson et al.

Here is an example of styling a photo of San Francisco with Van Gogh's Starry Night

The code is based off this paper by Johnson et al which in turn builds off of A Neural Algorithm of Artistic Style by Gatys et al.

This implementation uses Instance Norm described by Ulyanov et al and Resize-Convolution from Odena et al.

Takes a few hours to train on a P2 instance on AWS and image generation takes a few seconds on a Macbook Pro. Training image dataset was from MS COCO validation set and uses the VGG19 network for texture and style loss.

If you don't want to install tensorflow and deal with training your own models, try the Android app which has other styles trained using this code.

Get it on Google Play


  1. Tensorflow 0.12
  2. pip install:
  • scikit-image
  • numpy 1.11

Instructions for Processing

  1. Go to the project root (there is a pretrained model in the /data directory)
  2. Run:
$ python --input=path_to_image.jpg --output=your_output_file.jpg

Note: The script defaults to a pretrained model based off starry.jpg - if you wish to try a different style you will have to train your own in the next section

Intructions for Training

  1. Download VGG19 weights as vgg19.npy
  2. Download MS COCO validation dataset - the training script defaults to look for images in a directory named input_images, though you can add whatever directory you want as a command line argument
  3. Run:
$ python --data_dir=/path/to/ms_coco --texture=path/to/source_image.jpg

Instructions for Exporting

When training, your model gets saved as a series of tensorflow checkpoints - the default directory for this is the named 'model.' We export these checkpoints to a .pb file for easy usage, using the file copied from the tensorflow source repo. To run the export script run:

$ python --input_dir=path_to/training_directory
You can’t perform that action at this time.