Ilastik Feature Detection VS. FIJI AutoThreshold for DAPI stained nuclei?

I am just starting to test Ilastik for segmenting DAPI stained nuclei in confocal stacks (60X) 1 uM steps. I have been impressed with the ease of use of Ilastik, but I am not sure it is doing a better job than FIJI AutoThreshold set to “Triangle” thresholding. In my preliminary efforts Ilastik does not seem to outperform FIJI in simple cell nuclei identification and leaves many cells as clusters despite being given training during the Object Classification.

Curious if others have compared and have insights based on experience with this kind of problem.