Data Processing

This folder contains the following files:

• dataset.py: The Pytorch dataset used for training our model
• data_fetching.py: lenand get methods for each dataset
• data_utils.py: DataProcessing and PromptProcessing methods

Data Fetching

data_fetching.py implements lenand get methods for all used datasets. For this we assume our own file structure. More details in the section Our File Structure. However you can easily implement len and get for your own dataset. For this have a look at the section Using your own Cluster/Dataset.

Our File Structure

All of our data is stored in one data_directory. Each dataset has a subdirectory in this data_directory. We predefined the file structure for the helmholtz cluster. It is avilable with the keyword helmholtz in the cluster variable.

Using your own Cluster/Dataset

For existing datasets you can change the len and get methods in DATASET_DICT. For your own dataset you can implement your own len and get methods and add them to DATASET_DICT. len should return the number of samples in the dataset. get should return a tuple (image_path, mask_path) where image_path is the path to the image file and mask_path is the path to the mask file. You should also change the cluster variable to custom.

Dataset & Data Utils

You probably don't need to change anything in dataset.py and data_utils.py.