Alternatively you can build your own dataset by setting up the following directory structure:
.
├── data
| ├── edge_smooth_image
| | ├── image # Training
| | | ├── A # Contains domain edge smoothed images
| ├── target
| | ├── image # Training
| | | ├── A # Contains domain target style images
| ├── photo
| | ├── image # Training
| | | ├── A # Contains domain real world images
python train.py --checkpoint_dir=path/to/ckpt --model_config=path/to/model_config
following directory structure:
.
├── input_dir
| ├── image # Training
| | ├── A # Contains input images
python infer.py --weights=path/to/generator.pt --model_config=path/to/model_config --input=path/to/input --output_dir=path/to/output_dir