Focus alignment prior to training of the algorithm improves the F1-scorea.

aParameter combinations for the detection algorithm were optimised for all images in a given dataset (‘Training Set’). The F1-scores when the parameter combinations were applied to the same or other manually-labelled images (‘Test Set’) were averaged. Training was performed using the manually-detected foci (‘Raw’) or after the manually-selected foci were aligned to their appropriate local maximum (‘Aligned’). Values are shown to three significant figures.Focus alignment prior to training of the algorithm improves the F1-scorea.