Final submissions¶

Evaluation code (python) will be released so that participants can measure model performance on their own validation splits. Participants will also be provided with a template for creating the final submission file, consisting of model-predicted segmentation masks for each structure. Predicted segmentation masks on the test set are expected to be saved as hdf5 files and uploaded to an online portal (link will be shared with participants via email).

Self-evaluation¶

Participants are encouraged to split the training data into train/validation splits to evaluate their work prior to final submission and evaluation on the test data. No separate validation data will be provided.

Participants are allowed 3 total submissions and the model with the highest performing score will be considered.