A PyTorch implementation of The Conditional Analogy GAN combined with StackGAN-v2
Code based on the original implementation in keras
The following repositories were also referenced:

• https://github.com/hanzhanggit/StackGAN-v2
• https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix

• Python 3.6
• PyTorch 0.4.0
• Visdom

The dataset is just same as here and the folder of dataset is structured as follows:

├── data
    ├── imgs_test
    │   ├── 1
    │   └── 5
    └── imgs_train
        ├── 1
        └── 5

• Open a visdom server

python -m visdom.server

• Train a model

python main.py

Use trained model to generate images and calculate IS score

python main.py --mode test --model_dir logs/origin/model_weight