CSAW-S: mammography screenings from 172 different patients with annotations for semantic segmentation

To download the dataset follow this link

If you use this dataset please cite our work Adding seemingly uninformative labels helps in low data regimes

Reproducibility

To convert the dataset to the version we used for our experiments open the csaws_creation folder and run the script generate_dataset.py

Enviroment setup

To install the enviroment we use run: conda env create -f environment.yml

Data structure

current folder 
│ 
├── CsawS 
│     │
│     ├ anonymized_dataset -> patient folders with all the screenings and annotations (tumors - expert 1)
│     │
│     ├ test_data
│     │     │
│     │     ├ anonymized_dataset -> patient folders with all the screenings
│     │     │
│     │     ├ annotator_1 -> annotations from expert 1
│     │     │
│     │     ├ annotator_2 -> annotations from expert 2
│     │     │
│     │     └ annotator_3 -> annotations from expert 3
│     │
│     └ training_random_splits.json -> the splits we used for our experiments (can be ommited)
│ 
├── csaws_creation -> scripts to convert the dataset to the version we used for our experiments
│ 
└── README.md -> current file