Table 3.

The details of learning setting of the models.

App.	Loss	ILR	No. of Epochs	Batch Size	Input Size	Data Aug.
EM	DC + BCE 1	3 × 10−4	200	16	128 × 128	Yes
Nuclei	DC + BCE	3 × 10−4	100	16	128 × 128	Yes
Polyp	DC + WCE 2	3 × 10−4	200	8	224 × 224	Yes
Skin lesion	DC + BCE	3 × 10−4	200	8	224 × 224	Yes
Left atrium	DC + WCE	1× 10−3	100	16	128 × 128	Yes
Spleen	DC + WCE	3 × 10−4	200	8	224 × 224	Yes
Liver	DC + WCE	3 × 10−4	100	8	224 × 224	No

¹ Dice loss + binary cross-entropy; ² Dice loss + weighted cross-entropy.