Table 2 U-Net architecture overview.

From: Integrating radiomic texture analysis and deep learning for automated myocardial infarction detection in cine-MRI

Layer (type)

Output shape

Parameter

Input Layer

(None, 128, 128, 3)

0

Conv2D

(None, 128, 128, 32)

896

Conv2D

(None, 128, 128, 32)

9,248

Dropout

(None, 128, 128, 32)

0

MaxPooling2D

(None, 64, 64, 32)

0

Conv2D

(None, 64, 64, 64)

18,496

Conv2D

(None, 64, 64, 64)

36,928

Dropout

(None, 64, 64, 64)

0

MaxPooling2D

(None, 32, 32, 64)

0

Conv2D

(None, 32, 32, 128)

73,856

Conv2D

(None, 32, 32, 128)

147,584

Dropout

(None, 32, 32, 128)

0

Conv2D_Transpose

(None, 64, 64, 64)

32,832

Concatenate

(None, 64, 64, 128)

0

Conv2D

(None, 64, 64, 64)

73,792

Conv2D_Transpose

(None, 128, 128, 32)

8,224

Concatenate

(None, 128, 128, 64)

0

Conv2D

(None, 128, 128, 32)

18,464

Conv2D

(None, 128, 128, 1)

33