Pix2Pix Example by NNabla
NNabla example code of Image-to-Image Translation with Conditional Adversarial Nets a.k.a. pix2pix.
Result Examples (cherry picked)
Train Pix2Pix Model
python pix2pix.py --train -c "cudnn" -d 0 --path-gan
Trained model will be saved at log directory (default is
Generate Fakes from Trained Pix2Pix Model
python pix2pix.py --generate --model "<path to your trained model>" -c "cudnn" -d 0
Generated images will be saved at log directory (default is
Currently, this example is only for the facade dataset.
- P. Isola, J. Zhu, T. Zhou, and A. A. Efros. "Image-to-Image Translation with Conditional Adversarial Networks," arXiv:1611.07004, 2016.