Multichannel Retinal Blood Vessel Segmentation Based on the Combination of Matched Filter and U-Net Network

Table 8

Network comparison of the proposed method with other methods based on U-Net.

Method	Input size	Epoch	Training images	Parameters
Ronneberger et al. (2015) [24]		N/A	N/A	28.94 M
Yan et al. (2018) [13]		Over 30	26052 (4)	30.96 M
Jiang et al. (2018) [19]		30	42421 (4)	58.31 M
Soomro et al. (2019) [39]	Original image	N/A	3959 (4)	4.71 M
Joshua et al. (2020) [40]		50	113 (4)	0.64 M
Zhang et al. (2020) [17]		120	116 (3)	3.86 M
Budak et al. (2020) [34]		Over 30	Over 45840 (2)	0.97 M
Proposed method		30	220 (3)	13.39 M

means training images of datasets.