This repo is heavily based on Original CycleGAN implementation. Most of our work involves adding code to better handle the dataset we are working with.

• Linux or macOS
• Python 3
• CPU or NVIDIA GPU + CUDA CuDNN

• Clone this repo:

git clone https://github.com/lsh0357/Baby-Face-Generator-with-CycleGAN.git
cd Baby-Face-Generator-with-CycleGAN

• Install [PyTorch](http://pytorch.org and) 0.4+ and other dependencies (e.g., torchvision, visdom and dominate).
  • For pip users, please type the command pip install -r requirements.txt.

• If you would like to try the model with 2 or more input images, you could use http://www.morphthing.com/morph to morph your images, else if only one image as input, look into the next step.
• The model is restricted to face only images, so please follow https://github.com/kb22/Create-Face-Data-from-Images to extract face from images or use any other tool to clean up the background noisies.
• The trained model is saved at ./trained_model. Move the whole trained_model directory to ./checkpoints
• To generate a baby style face of your input, simply run

python test.py --dataroot <PATH_TO_TEST_SET> --name trained_model --model test

• Look at your baby face at ./results

• To view training results and loss plots, run python -m visdom.server and click the URL http://localhost:8097.
• Train a model:

#!./scripts/train_cyclegan.sh
python train.py --dataroot <PATH_TO_TRAINING_SET> --name <NAME_FOR_THE_MODEL> --model cycle_gan

To see more intermediate results, check out ./checkpoints/<NAME_FOR_THE_MODEL>.

• Test the model:

#!./scripts/test_cyclegan.sh
python test.py --dataroot <PATH_TO_TEST_SET> --name <NAME_FOR_THE_MODEL> --model cycle_gan
python test.py --dataroot <PATH_TO_TEST_SET> --name <NAME_FOR_THE_MODEL> --model test(For one side test, such as only from adult to baby)

• The test results will be saved to a html file here: ./results/<NAME_FOR_THE_MODEL>/latest_test/index.html.

• Look at the options directory or the original repo for more information about how to do a pre-train and pther advantage training and testing options.