/Deep-image-analogy-TensorFlow

Implementation of the paper : Deep image analogy

Primary LanguagePythonMIT LicenseMIT

Deep-image-analogy-TensorFlow

Implementation of the paper : Deep image analogy

Introduction

This code mainly implements the interesting paper Visual Attribute Transfer through Deep Image Analogy, details of the algorithm please see the paper.

How to use

Step 1. Download the vgg19.npy(BaiduDrive), and put it into the folder 'vgg19'

Step 2. Modify the image's path in Main.py

Step 3. execute Main.py

Python packages

===================

  1. python 3.5
  2. tensorflow1.4.0
  3. numpy
  4. scipy
  5. pillow

===================

Results

Input : Output :: Output : Input
: :: :
: :: :
: :: :

Time: 3 minutes for 224x224 image size

These results are not satisfactory compare to paper's attractive results

Acknowledgement

Thanks for the PatchMatch code from harveyslash