using the available trained models for fine-tuning over a new dataset

[nattaran]

I am going to fine-tune the available semantic segmentation models(in this repo). for fine-tuning I am going to use the Aeroscapes dataset which has both aerial images + masks. I don't know how to add the new data to be used for training. What folder structure is needed (train, valid)?
Please let me know if you used the available YMl files and models from this repo for fine-tuning?