superresolution_gan

Chainer v5.1 implementation of Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
Original source code was implemented by Hi-king
This source code was updated to work with Chainer v5.1

Training

python train.py \
  --gpu=0  \
  --pretrained_generator generator_model_3008000.npz \
  --src ./src
  --batchsize=16 \
  --epoch=10
  --k_mse=0.0001 \
  --k_adversarial=0.00001 \
  --outdirname output
  --vgg