https://www.kaggle.com/c/sartorius-cell-instance-segmentation/discussion/293159
I tried out the Cellpose. The model is based on U-Net, however rather than training it directly on bitmask targets they first convert them to "spatial flows" representations and train on that. This makes segmentation of dense and touching cells more reliable. See a sample output on the fourth image bellow.
It works well out of the box (without any custom code in training, nor inference) , and gives a better LB score than the other UNet models I've seen shared so far.