chainer_progressive_gan

Learning animeface latent space using Progressive GAN

Create Kawaii !

pip install git+https://github.com/Hi-king/chainer_progressive_gan.git
# need git-lfs(https://github.com/git-lfs/git-lfs/wiki/Installation)

from chainer_progressive_gan import KawaiiGenerator
creator = KawaiiGenerator()
image = creator.create_one()
image.save("test.png")

Re-training

python train.py --gpu=1 --resize 256 your/dataset/path

Result

Stage-by-Stage animation

Conditional Image Generation

sketch2img

python tools/conditionals/edge2img/predict_edge2img.py --input_image signico_face.png --vectorizer result/edge2img_resize256_stage0.0_batch16_stginterval500000_latentON_1538310505/vectorizer_280000.npz --generator result/edge2img_resize256_stage0.0_batch16_stginterval500000_latentON_1538310505/generator_280000.npz --stage 8 --to_line --use_latent

input	colorized

pose2img

imageA	poseA	poseB	imageB(generated)

Acknowledgements

• Progressive Growing of GANs for Improved Quality, Stability, and Variation
  • Tero Karras, Timo Aila, Samuli Laine, Jaakko Lehtinen
  • http://research.nvidia.com/publication/2017-10_Progressive-Growing-of
• chainer-gan-lib
  • https://github.com/pfnet-research/chainer-gan-lib
  • My implementation highly depends on this repository
• labpcascade_animeface
  • https://github.com/nagadomi/lbpcascade_animeface