We have tried two different choices to select best samples from prediction data volume: Gaussian similarity-mean filter and lower-upper-middle (LUM) filter.
![[pdf]](icons/pdf.png){kind=icon}