Code to synthesise textures using convolutional neural networks as described in Gatys et al. 2015 (
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
DeepImageSynthesis fixed resizing of mean file Jan 5, 2016
Images initial commit Oct 20, 2015
Models initial commit Oct 20, 2015
Example.ipynb fixed typo in name of model file Mar 7, 2016 Update Nov 3, 2015


Code to synthesise textures using convolutional neural networks as described in the paper "Texture Synthesis Using Convolutional Neural Networks" (Gatys et al., NIPS 2015) ( More examples of synthesised textures can be found at

The IPythonNotebook Example.ipynb contains the code to synthesise the pebble texture shown in Figure 3A (177k parameters) of the revised version of the paper. In the notebook I additionally match the pixel histograms in each colorchannel of the synthesised and original texture, which is not done in the figures in the paper. #Prerequisites

  • To run the code you need a recent version of the Caffe deep learning framework and its dependencies (tested with master branch at commit 20c474fe40fe43dee68545dc80809f30ccdbf99b).
  • The images in the paper were generated using a normalised version of the 19-layer VGG-Network described in the work by Simonyan and Zisserman. The weights in the normalised network are scaled such that the mean activation of each filter over images and positions is equal to 1. The normalised network can be downloaded here and has to be copied into the Models/ folder.


This software is published for academic and non-commercial use only.