Aerial Imagery with .TIF Extension 

Geometry Masks in .geojson Format 

SAM Model Checkpoint downloaded from https://dl.fbaipublicfiles.com/segment_anything/sam_vit_b_01ec64.pth

Put the downloaded checkpoint into pretrained/ folder

Go to preproccessing Folder and Create python environment using requirements.txt (Preprocessing uses different set of packages than trainig and must be done only once) 

After Activating the newly created Environment run: 
python run --tif-folder $ABSOLUTE_PATH_TO_WHERE_TIF_IMAGES_RESIDE --geojson-folder $ABSOLUTE_PATH_WHERE_GEOMETRY_MASKS_RESIDE

This will create and put resutls into ./temp folder
copy load folder from ./temp and put it into the root workspace path
load folder contains everython you need for training, validation, and testing 

Go to workspace Folder and Create python environment using requirements.txt
Activate that workspace and run:

torchrun train.py --config configs/sam-vit-b.yaml

python test.py --model $ABSOLUTE_PATH_TO_THE_TRAINED_MODEL ----dataset ('val_dataset'|'test_dataset')