Table 1.

The number of images and cells in each dataset split.

Data split	No. of image	R	G	U
Train	967	22800	24862	24575
Validation	120	2823	2972	3135
Test	120	3336	(3374)	(3459)

Numbers in brackets were estimated from model prediction results under the guidance of fluorescence signal. For reference, the average precision for R, G, and U class on the validation split is 89.2, 88.0, and 80.8, respectively.