Unofficial implementation for paper "AgileGAN: Stylizing Portraits by Inversion-Consistent Transfer Learning"
The code is mainly based on these two repo: AgileGAN-inference and pixel2style2pixel
Please refer to AgileGAN#requirements and pixel2style2pixel#prerequisites.
-
Prepare the dataset and pretrained models.
-
Start training using
python train.py \
--exp_dir=ffhq_encode \
--workers=8 \
--batch_size=4 \
--test_batch_size=4 \
--test_workers=8 \
--val_interval=2500 \
--save_interval=5000 \
--start_from_latent_avg
Please refer to AgileGAN-inference
python test.py --path examples/29899.jpg --style cartoon