Defocusing makes DeepMoMA loose segment hypothesis

[TheoGervais]

In general I can really see on this dataset that defocusing makes deep MoMA trip out : temporary defocus, e.g. between t=344 and t=354 on Pos1 are accompanied by "loss" of the cell detection (no segment hypothesis anymore) at the top of the GL. The focus then comes back, and suddenly all segment hypothesis are appearing again. I have several example of this in this dataset if you need more.

Here on the central frame the two upper cells do not correspond to any segment hypothesis. It's often worse than that, and can be problematic since the segments hypothesis come and go, which means that the user has to deal with "appearing and disappearing"cells.