search

linfengWen98 / CAP-VSTNet

[CVPR 2023] CAP-VSTNet: Content Affinity Preserved Versatile Style Transfer
MIT License
120 stars 8 forks source link

Optimize the effect of style transfer #7

Open Zachary-ZCH opened 1 year ago

Zachary-ZCH commented 1 year ago

Thank you for your excellent work. Recently, when trying to transfer the style of smartphone photos to camera style, I found that because the background in DSLR photos has a blurred effect, so the transferred smartphone photos also became blurred. Is there any way to optimize this?

linfengWen98 commented 12 months ago

Try running train.py with --win_rad 2. Hope it helps. Note: The larger the value, the more detail in the generated image and the higher the CPU and memory requirements (proportional to the win_rad**2).