Lack of Critical Evaluation After Cell Segmentation

[DarioS]

InstanSeg (the only one with no I.M.C. data in its training set) appears to be missing lots of cells that the other two easily detect.

Also, how should a user decide on the correct segmentation when algorithms disagree with each other? Any ground truth?