taejangpark / stylegan

StyleGAN

This is an implementation of a Photo2Style Generator Architecture for GAN on tensorflow. The model generates new image with refernce style or vice versa.

Getting Started

• (P2SNet.py, config.py): These files are the main Photo to Style Generator network.

• (datapipe.py): This file's role is loading and changing to tensor your dataset taht are in photo2rtyle folder. you must put your training dataset in (photo2style)

• (pretrained_models): you put the pretrained (VGG 19) model in this folder.

• (train.py): this file is for training. To train this network ''' python train.py '''

• (inference.py): this file is for inference. we have 2 inference mode that are reference style and random style and you set True or False on --rand_style value. To run this file ''' python infernece.py --rand_style=True'''

Result

Photo	Style
Photo with style	Style with Photo
Photo with rand style	Style with rand Photo