U-Net segmentation as presented in paper Caicedo, J. C., et. al., Evaluation of Deep Learning Strategies for Nucleus Segmentation in Fluorescence Images, Cytometry Part A, 95(9), doi: 10.1002/cyto.a.23863
Model predicts three classes: background, edge and foreground. The model was trained with DSB 2018 training set: https://www.kaggle.com/c/data-science-bowl-2018