Personal implementation of paper PairedCycleGAN: Asymmetric Style Transfer for Applying and Removing Makeup
Please organize your training/testing dataset like:
├─ dataroot
│ ├─ sketch (real A)
│ ├─ reference (real B)
Or you can change the dataset code (in datasets/dataset.py) to fit your training dataset.
introduced in the paper is not adopted and implemented according to my task.
Basic training command:
python3 train.py --name [your_project_name] --dataroot [your_training_dataset_path]
See options/options.py for more information regarding the command.
Checkpoint files is saved in ./checkpoints/[project_name] in default settings.