Skip to content

Latest commit

 

History

History
71 lines (47 loc) · 2.92 KB

USAGE.md

File metadata and controls

71 lines (47 loc) · 2.92 KB

License

Copyright (C) 2018 NVIDIA Corporation. All rights reserved. Licensed under the CC BY-NC-SA 4.0 license (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).

Preparation 1: Setup environment and install required libraries

  • OS: Ubuntu 16.04
  • CUDA: 9.1
  • Python: Python 2 from Anaconda2
  • Python Library Dependency
    1. conda install pytorch torchvision cuda90 -y -c pytorch
    2. conda install -y -c menpo opencv3
    3. conda install -y -c anaconda pip
    4. pip install scikit-umfpack
    5. pip install cupy
    6. pip install pynvrtc

Preparation 2: Download pretrained networks

  • Download pretrained networks via the following link.
  • Unzip and store the model files under models.

Example 1: Transfer the style of a style photo to a content photo.

  • Go to the image folder: cd images

  • Download content image 1: axel -n 1 http://freebigpictures.com/wp-content/uploads/shady-forest.jpg --output=content1.png

  • Download style image 1: axel -n 1 https://vignette.wikia.nocookie.net/strangerthings8338/images/e/e0/Wiki-background.jpeg/revision/latest?cb=20170522192233 --output=style1.png

  • These images are huge. We need to resize them first. Run

    • convert -resize 25% content1.png content1.png
    • convert -resize 50% style1.png style1.png

  • Go to the root folder: cd ..

  • Content image

  • Style image

  • Test the photorealistic image stylization code python demo.py

  • Stylized content image

TODO Example 2: Transfer the style of a style photo to a content photo with semantic label maps.

Label map

By default, our algorithm performs the global stylization. In order to give users control to decide the content–style correspondences for better stylization effects, we also support the spatial control through manully drawing label maps.

  • Install the tool labelme and run the following command to start it:
labelme

  • Start labeling regions in the content& style image. The corresponding regions (e.g., sky-to-sky) should have the same label.

  • The labeling result is saved in a ".json" file. By running the following command, you will get the "label.png" under "path/example_json", which is the label map used in our code.

labelme_json_to_dataset example.json -o path/example_json

"label.png" is a 1-channel image (usually looks totally black) consists of consecutive labels starting from 0. You will get a visualized reuslt "label_viz.png" at the same time.

Acknowledgement

  • We express gratitudes to the great work DPST by Luan et al. and their Torch and Tensorflow implementations.