from Nested_UNet import Nested_UNet
model = Nested_UNet(channel=1, mode='accurate')
- Implemented UNet++ (Nested UNet) through PyTorch.
- Re-designed skip pathways are successfully implemented, and Deep supervision is provided by hyperparameter to enable the mode of Deep supervision.
- The mode provides Accurate Mode, Fast Mode, as suggested in the paper.
if mode == 'accurate':
self.deep_supervision = True
elif mode == 'fast':
self.deep_supervision = False
else:
raise ValueError("check 'mode' arguement")
- PyTorch를 통해 UNet++(Nested UNet)를 구현하였습니다.
- Re-designed skip pathways를 성공적으로 구현하며, Deep supervision의 모드를 설정할 수 있게끔 hyperparameter로 제공합니다.
- 모드는 논문에서 제안되었듯이 Accurate Mode, Fast Mode를 제공합니다.
if mode == 'accurate':
self.deep_supervision = True
elif mode == 'fast':
self.deep_supervision = False
else:
raise ValueError("check 'mode' arguement")
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 32, 224, 224] 320
BatchNorm2d-2 [-1, 32, 224, 224] 64
ReLU-3 [-1, 32, 224, 224] 0
Conv2d-4 [-1, 32, 224, 224] 9,248
BatchNorm2d-5 [-1, 32, 224, 224] 64
ReLU-6 [-1, 32, 224, 224] 0
ConvBlock-7 [-1, 32, 224, 224] 0
MaxPool2d-8 [-1, 32, 112, 112] 0
EncoderBlock-9 [[-1, 32, 224, 224], [-1, 32, 112, 112]] 0
Conv2d-10 [-1, 64, 112, 112] 18,496
BatchNorm2d-11 [-1, 64, 112, 112] 128
ReLU-12 [-1, 64, 112, 112] 0
Conv2d-13 [-1, 64, 112, 112] 36,928
BatchNorm2d-14 [-1, 64, 112, 112] 128
ReLU-15 [-1, 64, 112, 112] 0
ConvBlock-16 [-1, 64, 112, 112] 0
MaxPool2d-17 [-1, 64, 56, 56] 0
EncoderBlock-18 [[-1, 64, 112, 112], [-1, 64, 56, 56]] 0
Conv2d-19 [-1, 128, 56, 56] 73,856
BatchNorm2d-20 [-1, 128, 56, 56] 256
ReLU-21 [-1, 128, 56, 56] 0
Conv2d-22 [-1, 128, 56, 56] 147,584
BatchNorm2d-23 [-1, 128, 56, 56] 256
ReLU-24 [-1, 128, 56, 56] 0
ConvBlock-25 [-1, 128, 56, 56] 0
MaxPool2d-26 [-1, 128, 28, 28] 0
EncoderBlock-27 [[-1, 128, 56, 56], [-1, 128, 28, 28]] 0
Conv2d-28 [-1, 256, 28, 28] 295,168
BatchNorm2d-29 [-1, 256, 28, 28] 512
ReLU-30 [-1, 256, 28, 28] 0
Conv2d-31 [-1, 256, 28, 28] 590,080
BatchNorm2d-32 [-1, 256, 28, 28] 512
ReLU-33 [-1, 256, 28, 28] 0
ConvBlock-34 [-1, 256, 28, 28] 0
MaxPool2d-35 [-1, 256, 14, 14] 0
EncoderBlock-36 [[-1, 256, 28, 28], [-1, 256, 14, 14]] 0
Conv2d-37 [-1, 512, 14, 14] 1,180,160
BatchNorm2d-38 [-1, 512, 14, 14] 1,024
ReLU-39 [-1, 512, 14, 14] 0
Conv2d-40 [-1, 512, 14, 14] 2,359,808
BatchNorm2d-41 [-1, 512, 14, 14] 1,024
ReLU-42 [-1, 512, 14, 14] 0
ConvBlock-43 [-1, 512, 14, 14] 0
MaxPool2d-44 [-1, 512, 7, 7] 0
EncoderBlock-45 [[-1, 512, 14, 14], [-1, 512, 7, 7]] 0
ConvTranspose2d-46 [-1, 32, 224, 224] 8,224
Conv2d-47 [-1, 32, 224, 224] 18,464
BatchNorm2d-48 [-1, 32, 224, 224] 64
ReLU-49 [-1, 32, 224, 224] 0
Conv2d-50 [-1, 32, 224, 224] 9,248
BatchNorm2d-51 [-1, 32, 224, 224] 64
ReLU-52 [-1, 32, 224, 224] 0
ConvBlock-53 [-1, 32, 224, 224] 0
DecoderBlock-54 [-1, 32, 224, 224] 0
ConvTranspose2d-55 [-1, 64, 112, 112] 32,832
Conv2d-56 [-1, 64, 112, 112] 73,792
BatchNorm2d-57 [-1, 64, 112, 112] 128
ReLU-58 [-1, 64, 112, 112] 0
Conv2d-59 [-1, 64, 112, 112] 36,928
BatchNorm2d-60 [-1, 64, 112, 112] 128
ReLU-61 [-1, 64, 112, 112] 0
ConvBlock-62 [-1, 64, 112, 112] 0
DecoderBlock-63 [-1, 64, 112, 112] 0
ConvTranspose2d-64 [-1, 128, 56, 56] 131,200
Conv2d-65 [-1, 128, 56, 56] 295,040
BatchNorm2d-66 [-1, 128, 56, 56] 256
ReLU-67 [-1, 128, 56, 56] 0
Conv2d-68 [-1, 128, 56, 56] 147,584
BatchNorm2d-69 [-1, 128, 56, 56] 256
ReLU-70 [-1, 128, 56, 56] 0
ConvBlock-71 [-1, 128, 56, 56] 0
DecoderBlock-72 [-1, 128, 56, 56] 0
ConvTranspose2d-73 [-1, 256, 28, 28] 524,544
Conv2d-74 [-1, 256, 28, 28] 1,179,904
BatchNorm2d-75 [-1, 256, 28, 28] 512
ReLU-76 [-1, 256, 28, 28] 0
Conv2d-77 [-1, 256, 28, 28] 590,080
BatchNorm2d-78 [-1, 256, 28, 28] 512
ReLU-79 [-1, 256, 28, 28] 0
ConvBlock-80 [-1, 256, 28, 28] 0
DecoderBlock-81 [-1, 256, 28, 28] 0
ConvTranspose2d-82 [-1, 32, 224, 224] 8,224
Conv2d-83 [-1, 32, 224, 224] 27,680
BatchNorm2d-84 [-1, 32, 224, 224] 64
ReLU-85 [-1, 32, 224, 224] 0
Conv2d-86 [-1, 32, 224, 224] 9,248
BatchNorm2d-87 [-1, 32, 224, 224] 64
ReLU-88 [-1, 32, 224, 224] 0
ConvBlock-89 [-1, 32, 224, 224] 0
DecoderBlock-90 [-1, 32, 224, 224] 0
ConvTranspose2d-91 [-1, 64, 112, 112] 32,832
Conv2d-92 [-1, 64, 112, 112] 110,656
BatchNorm2d-93 [-1, 64, 112, 112] 128
ReLU-94 [-1, 64, 112, 112] 0
Conv2d-95 [-1, 64, 112, 112] 36,928
BatchNorm2d-96 [-1, 64, 112, 112] 128
ReLU-97 [-1, 64, 112, 112] 0
ConvBlock-98 [-1, 64, 112, 112] 0
DecoderBlock-99 [-1, 64, 112, 112] 0
ConvTranspose2d-100 [-1, 128, 56, 56] 131,200
Conv2d-101 [-1, 128, 56, 56] 442,496
BatchNorm2d-102 [-1, 128, 56, 56] 256
ReLU-103 [-1, 128, 56, 56] 0
Conv2d-104 [-1, 128, 56, 56] 147,584
BatchNorm2d-105 [-1, 128, 56, 56] 256
ReLU-106 [-1, 128, 56, 56] 0
ConvBlock-107 [-1, 128, 56, 56] 0
DecoderBlock-108 [-1, 128, 56, 56] 0
ConvTranspose2d-109 [-1, 32, 224, 224] 8,224
Conv2d-110 [-1, 32, 224, 224] 36,896
BatchNorm2d-111 [-1, 32, 224, 224] 64
ReLU-112 [-1, 32, 224, 224] 0
Conv2d-113 [-1, 32, 224, 224] 9,248
BatchNorm2d-114 [-1, 32, 224, 224] 64
ReLU-115 [-1, 32, 224, 224] 0
ConvBlock-116 [-1, 32, 224, 224] 0
DecoderBlock-117 [-1, 32, 224, 224] 0
ConvTranspose2d-118 [-1, 64, 112, 112] 32,832
Conv2d-119 [-1, 64, 112, 112] 147,520
BatchNorm2d-120 [-1, 64, 112, 112] 128
ReLU-121 [-1, 64, 112, 112] 0
Conv2d-122 [-1, 64, 112, 112] 36,928
BatchNorm2d-123 [-1, 64, 112, 112] 128
ReLU-124 [-1, 64, 112, 112] 0
ConvBlock-125 [-1, 64, 112, 112] 0
DecoderBlock-126 [-1, 64, 112, 112] 0
ConvTranspose2d-127 [-1, 32, 224, 224] 8,224
Conv2d-128 [-1, 32, 224, 224] 46,112
BatchNorm2d-129 [-1, 32, 224, 224] 64
ReLU-130 [-1, 32, 224, 224] 0
Conv2d-131 [-1, 32, 224, 224] 9,248
BatchNorm2d-132 [-1, 32, 224, 224] 64
ReLU-133 [-1, 32, 224, 224] 0
ConvBlock-134 [-1, 32, 224, 224] 0
DecoderBlock-135 [-1, 32, 224, 224] 0
Conv2d-136 [-1, 1, 224, 224] 33
Conv2d-137 [-1, 1, 224, 224] 33
Conv2d-138 [-1, 1, 224, 224] 33
Conv2d-139 [-1, 1, 224, 224] 33
Sigmoid-140 [-1, 1, 224, 224] 0
================================================================
Total params: 9,048,996
Trainable params: 9,048,996
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.19
Forward/backward pass size (MB): 4521.27
Params size (MB): 34.52
Estimated Total Size (MB): 4555.98
----------------------------------------------------------------
Paper 1: UNet++: A Nested U-Net Architecture for Medical Image Segmentation