Usage for mosaic:

1. You have to create directory 'trains' which will contains 3 mode directories (nnUNet_preprocessed, nnUNet_raw, nnUNet_results).

The folder structure should be like this:

└── trains
    ├── nnUNet_preprocessed
    ├── nnUNet_raw
    └── nnUNet_results

2. You have to set the env variables.

For example it should be like:

export nnUNet_raw="/bluemind/nnunet/trains/nnUNet_raw"

export nnUNet_preprocessed="/bluemind/nnunet/trains/nnUNet_preprocessed"

export nnUNet_results="/bluemind/nnunet/trains/nnUNet_results"

3. You have to put your dataset into nnUNet_raw directory and then create dataset.json.

Use generate_json for this.

4. Use scripts for training.

There are several scripts written by me (Roma) for datgaset preprocessing, training, finetuning.

READ NNUUNET README FOR MORE INFORMATION