rbqc pipeline is fully built. it took hours to get this done with so many spontaneously thought out logic from my end. i added/enforced a lot of rules that i don't want to go too in depth here since there's a 4000-paragraph .md file detailing everything. it looks good. really good. im exhausted. been coding from 1pm-ish to 1am.
the anabsis of this batch will see more sacrifices. column 12 should be dropped for statistical significance and reduction of physical noise. physics and biology should be separated. and since ground truth is biology, physics-influenced changes should be discarded as such. we are down from 93 to 85. and this shall be discussed further with b tmr perhaps.
with current level of sophistication for the segmentation pipeline almost all previous masks are fixed for biological significance.
hull-dependent features are dropped entirely. we are left with circularity, area, and eccentricity. all three are either semi or completely independent from each other during calculations. manually-enforced regularizations.
is this good or bad? im not sure. we will wait to see.