Project: https://www.kaggle.com/competitions/data-science-bowl-2018/
We may need to take a subset of the training data. We also still need to think about and perform a train-validate split. Also, as it's a Kaggle competition, you do not have access to the test set labels, but you can still make "late submissions" to see your test score.
One solution in which you designed and tuned your own network and at least one solution which uses transfer learning for the encoder (see https://www.tensorflow.org/tutorials/images/segmentation for an example).