Li2017Universal

This is a PyTorch implementation for the paper "Universal Style Transfer via Feature Transforms", which comes from the repository PytorchWCT.

Test (Ultra-high Resolution Style Transfer)

Use --content and --style to provide the respective path to the content and style image.

CUDA_VISIBLE_DEVICES=<gpu_id> python test.py --content <content_path> \
                                             --style <style_path> \
                                             --URST

For example:

CUDA_VISIBLE_DEVICES=0 python test.py --content ../examples/content/pexels-andrea-piacquadio-3830880.jpg \
                                      --style ../examples/style/line2.png \
                                      --padding 64
                                      --URST

Some options:

• --patch_size: The maximum size of each patch. The default setting is 1000.
• --style_size: The size of the style image. The default setting is 1024.
• --thumb_size: The size of the thumbnail image. The default setting is 1024.
• --URST: Use our URST framework to process ultra-high resolution images.
• --alpha: Adjust the degree of stylization. It should be a value between 0.0 and 1.0 (default).

For more details please refer to PytorchWCT.