modules.resnet.DecoderNet2d¶
net = mdnc.modules.resnet.DecoderNet2d(
channel, layers, out_size, block='bottleneck',
kernel_size=3, in_length=2, out_planes=1
)
This moule is a built-in model for 2D residual decoder network. This network could be used as a part of the auto-encoder, or just a network for up-sampling (or generating) data.
The network would up-sample the input data according to the network depth. The depth is given by the length of the argument layers. The network structure is shown in the following chart:
flowchart TB
u1["Block 1<br>Stack of layers[0] blocks"]
u2["Block 2<br>Stack of layers[1] blocks"]
ui["Block ...<br>Stack of ... blocks"]
un["Block n<br>Stack of layers[n-1] blocks"]
optional:::blockoptional
subgraph optional [Optional]
cin["Conv2d<br>with unsqueeze"]
end
u1 -->|up<br>sampling| u2 -->|up<br>sampling| ui -->|up<br>sampling| un
cin -.-> u1
linkStyle 0,1,2 stroke-width:4px, stroke:#080 ;
classDef blockoptional fill:none, stroke-dasharray:10,10, stroke:#9370DB, width:100;The argument layers is a sequence of int. For each block \(i\), it contains layers[i-1] repeated residual blocks (see mdnc.modules.resnet.BlockPlain2d and mdnc.modules.resnet.BlockBottleneck2d). Each down-sampling or up-sampling is configured by stride=2. The channel number would be doubled in the up-sampling route. An optional unsqueezer and convolutional layer could be prepended to the first layer when the argument in_length != None. This optional layer is used for converting the vector features in initial feature maps.
Arguments¶
Requries
| Argument | Type | Description |
|---|---|---|
channel | int | The channel number of the first hidden block (layer). After each down-sampling, the channel number would be doubled. |
layers | (int,) | A sequence of layer numbers for each block. Each number represents the number of residual blocks of a stage (block). The stage numer, i.e. the depth of the network is the length of this list. |
out_size | int or(int, int) | The size of the output data. This argument needs to be specified by users, because the network needs to configure its layers according to the output size. |
block | str | The residual block type, could be:
|
kernel_size | int or(int, int) | The kernel size of each residual block. |
in_length | int | The length of the input vector, if not set, the input needs to be feature maps. See the property input_size to check the input data size in this case. |
out_planes | int | The channel number of the output data. |
Operators¶
__call__¶
y = net(x)
The forward operator implemented by the forward() method. The input data is a tensor with a size determined by configurations. The output is a 2D tensor. The channel number of the output is specified by the argument out_planes.
Requries
| Argument | Type | Description |
|---|---|---|
x | torch.Tensor | A tensor,
|
Returns
| Argument | Description |
|---|---|
y | A 2D tensor, the size should be (B, C, L1, L2), where B is the batch size, C is the input channel number, and (L1, L2) is the output data size specified by the argument out_size. |
Properties¶
nlayers¶
net.nlayers
The total number of convolutional layers along the depth of the network. This value would not take the fully-connected layer into consideration.
input_size¶
net.input_size
The size of the input data size (a tuple). This property is useful when in_length is None. In this case, the input size is determined by the network.
Warning
This size contains the channel number (as the first element), because the input channel number is also determined by network when in_length is None.
Examples¶
Example 1
1 2 3 4 5 6 | |
The number of convolutional layers along the depth is 33.
The input size is (32,).
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 1024, 2, 2] 132,096
Conv2d-2 [-1, 1024, 2, 2] 9,437,184
InstanceNorm2d-3 [-1, 1024, 2, 2] 2,048
PReLU-4 [-1, 1024, 2, 2] 1,024
Conv2d-5 [-1, 1024, 2, 2] 1,048,576
InstanceNorm2d-6 [-1, 1024, 2, 2] 2,048
PReLU-7 [-1, 1024, 2, 2] 1,024
Conv2d-8 [-1, 1024, 2, 2] 9,437,184
InstanceNorm2d-9 [-1, 1024, 2, 2] 2,048
PReLU-10 [-1, 1024, 2, 2] 1,024
Conv2d-11 [-1, 512, 2, 2] 524,288
Conv2d-12 [-1, 512, 2, 2] 524,288
InstanceNorm2d-13 [-1, 512, 2, 2] 1,024
_BlockBo...neckNd-14 [-1, 512, 2, 2] 0
InstanceNorm2d-15 [-1, 512, 2, 2] 1,024
PReLU-16 [-1, 512, 2, 2] 512
Conv2d-17 [-1, 512, 2, 2] 262,144
InstanceNorm2d-18 [-1, 512, 2, 2] 1,024
PReLU-19 [-1, 512, 2, 2] 512
Upsample-20 [-1, 512, 4, 4] 0
Conv2d-21 [-1, 512, 4, 4] 2,359,296
InstanceNorm2d-22 [-1, 512, 4, 4] 1,024
PReLU-23 [-1, 512, 4, 4] 512
Conv2d-24 [-1, 512, 4, 4] 262,144
Upsample-25 [-1, 512, 4, 4] 0
Conv2d-26 [-1, 512, 4, 4] 262,144
InstanceNorm2d-27 [-1, 512, 4, 4] 1,024
_BlockBo...neckNd-28 [-1, 512, 4, 4] 0
_BlockResStkNd-29 [-1, 512, 4, 4] 0
InstanceNorm2d-30 [-1, 512, 4, 4] 1,024
PReLU-31 [-1, 512, 4, 4] 512
Conv2d-32 [-1, 512, 4, 4] 262,144
InstanceNorm2d-33 [-1, 512, 4, 4] 1,024
PReLU-34 [-1, 512, 4, 4] 512
Conv2d-35 [-1, 512, 4, 4] 2,359,296
InstanceNorm2d-36 [-1, 512, 4, 4] 1,024
PReLU-37 [-1, 512, 4, 4] 512
Conv2d-38 [-1, 256, 4, 4] 131,072
Conv2d-39 [-1, 256, 4, 4] 131,072
InstanceNorm2d-40 [-1, 256, 4, 4] 512
_BlockBo...neckNd-41 [-1, 256, 4, 4] 0
InstanceNorm2d-42 [-1, 256, 4, 4] 512
PReLU-43 [-1, 256, 4, 4] 256
Conv2d-44 [-1, 256, 4, 4] 65,536
InstanceNorm2d-45 [-1, 256, 4, 4] 512
PReLU-46 [-1, 256, 4, 4] 256
Upsample-47 [-1, 256, 8, 8] 0
Conv2d-48 [-1, 256, 8, 8] 589,824
InstanceNorm2d-49 [-1, 256, 8, 8] 512
PReLU-50 [-1, 256, 8, 8] 256
Conv2d-51 [-1, 256, 8, 8] 65,536
Upsample-52 [-1, 256, 8, 8] 0
Conv2d-53 [-1, 256, 8, 8] 65,536
InstanceNorm2d-54 [-1, 256, 8, 8] 512
_BlockBo...neckNd-55 [-1, 256, 8, 8] 0
_BlockResStkNd-56 [-1, 256, 8, 8] 0
InstanceNorm2d-57 [-1, 256, 8, 8] 512
PReLU-58 [-1, 256, 8, 8] 256
Conv2d-59 [-1, 256, 8, 8] 65,536
InstanceNorm2d-60 [-1, 256, 8, 8] 512
PReLU-61 [-1, 256, 8, 8] 256
Conv2d-62 [-1, 256, 8, 8] 589,824
InstanceNorm2d-63 [-1, 256, 8, 8] 512
PReLU-64 [-1, 256, 8, 8] 256
Conv2d-65 [-1, 128, 8, 8] 32,768
Conv2d-66 [-1, 128, 8, 8] 32,768
InstanceNorm2d-67 [-1, 128, 8, 8] 256
_BlockBo...neckNd-68 [-1, 128, 8, 8] 0
InstanceNorm2d-69 [-1, 128, 8, 8] 256
PReLU-70 [-1, 128, 8, 8] 128
Conv2d-71 [-1, 128, 8, 8] 16,384
InstanceNorm2d-72 [-1, 128, 8, 8] 256
PReLU-73 [-1, 128, 8, 8] 128
Upsample-74 [-1, 128, 16, 16] 0
Conv2d-75 [-1, 128, 16, 16] 147,456
InstanceNorm2d-76 [-1, 128, 16, 16] 256
PReLU-77 [-1, 128, 16, 16] 128
Conv2d-78 [-1, 128, 16, 16] 16,384
Upsample-79 [-1, 128, 16, 16] 0
Conv2d-80 [-1, 128, 16, 16] 16,384
InstanceNorm2d-81 [-1, 128, 16, 16] 256
_BlockBo...neckNd-82 [-1, 128, 16, 16] 0
_BlockResStkNd-83 [-1, 128, 16, 16] 0
InstanceNorm2d-84 [-1, 128, 16, 16] 256
PReLU-85 [-1, 128, 16, 16] 128
Conv2d-86 [-1, 128, 16, 16] 16,384
InstanceNorm2d-87 [-1, 128, 16, 16] 256
PReLU-88 [-1, 128, 16, 16] 128
Conv2d-89 [-1, 128, 16, 16] 147,456
InstanceNorm2d-90 [-1, 128, 16, 16] 256
PReLU-91 [-1, 128, 16, 16] 128
Conv2d-92 [-1, 64, 16, 16] 8,192
Conv2d-93 [-1, 64, 16, 16] 8,192
InstanceNorm2d-94 [-1, 64, 16, 16] 128
_BlockBo...neckNd-95 [-1, 64, 16, 16] 0
InstanceNorm2d-96 [-1, 64, 16, 16] 128
PReLU-97 [-1, 64, 16, 16] 64
Conv2d-98 [-1, 64, 16, 16] 4,096
InstanceNorm2d-99 [-1, 64, 16, 16] 128
PReLU-100 [-1, 64, 16, 16] 64
Upsample-101 [-1, 64, 32, 32] 0
Conv2d-102 [-1, 64, 32, 32] 36,864
InstanceNorm2d-103 [-1, 64, 32, 32] 128
PReLU-104 [-1, 64, 32, 32] 64
Conv2d-105 [-1, 64, 32, 32] 4,096
Upsample-106 [-1, 64, 32, 32] 0
Conv2d-107 [-1, 64, 32, 32] 4,096
InstanceNorm2d-108 [-1, 64, 32, 32] 128
_BlockBo...eckNd-109 [-1, 64, 32, 32] 0
_BlockResStkNd-110 [-1, 64, 32, 32] 0
InstanceNorm2d-111 [-1, 64, 32, 32] 128
PReLU-112 [-1, 64, 32, 32] 64
Conv2d-113 [-1, 64, 32, 32] 4,096
InstanceNorm2d-114 [-1, 64, 32, 32] 128
PReLU-115 [-1, 64, 32, 32] 64
Conv2d-116 [-1, 64, 32, 32] 36,864
InstanceNorm2d-117 [-1, 64, 32, 32] 128
PReLU-118 [-1, 64, 32, 32] 64
Conv2d-119 [-1, 64, 32, 32] 4,096
_BlockBo...eckNd-120 [-1, 64, 32, 32] 0
InstanceNorm2d-121 [-1, 64, 32, 32] 128
PReLU-122 [-1, 64, 32, 32] 64
Conv2d-123 [-1, 64, 32, 32] 4,096
InstanceNorm2d-124 [-1, 64, 32, 32] 128
PReLU-125 [-1, 64, 32, 32] 64
Upsample-126 [-1, 64, 64, 64] 0
Conv2d-127 [-1, 64, 64, 64] 36,864
InstanceNorm2d-128 [-1, 64, 64, 64] 128
PReLU-129 [-1, 64, 64, 64] 64
Conv2d-130 [-1, 64, 64, 64] 4,096
Upsample-131 [-1, 64, 64, 64] 0
Conv2d-132 [-1, 64, 64, 64] 4,096
InstanceNorm2d-133 [-1, 64, 64, 64] 128
_BlockBo...eckNd-134 [-1, 64, 64, 64] 0
_BlockResStkNd-135 [-1, 64, 64, 64] 0
Conv2d-136 [-1, 3, 64, 63] 4,803
DecoderNet2d-137 [-1, 3, 64, 63] 0
================================================================
Total params: 29,196,291
Trainable params: 29,196,291
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.00
Forward/backward pass size (MB): 42.98
Params size (MB): 111.38
Estimated Total Size (MB): 154.36
----------------------------------------------------------------
Example 2
1 2 3 4 5 6 | |
The number of convolutional layers along the depth is 32.
The input size is (1024, 2, 2).
----------------------------------------------------------------
Layer (type) Output Shape Param #
================================================================
Conv2d-1 [-1, 1024, 2, 2] 9,437,184
InstanceNorm2d-2 [-1, 1024, 2, 2] 2,048
PReLU-3 [-1, 1024, 2, 2] 1,024
Conv2d-4 [-1, 1024, 2, 2] 1,048,576
InstanceNorm2d-5 [-1, 1024, 2, 2] 2,048
PReLU-6 [-1, 1024, 2, 2] 1,024
Conv2d-7 [-1, 1024, 2, 2] 9,437,184
InstanceNorm2d-8 [-1, 1024, 2, 2] 2,048
PReLU-9 [-1, 1024, 2, 2] 1,024
Conv2d-10 [-1, 512, 2, 2] 524,288
Conv2d-11 [-1, 512, 2, 2] 524,288
InstanceNorm2d-12 [-1, 512, 2, 2] 1,024
_BlockBo...neckNd-13 [-1, 512, 2, 2] 0
InstanceNorm2d-14 [-1, 512, 2, 2] 1,024
PReLU-15 [-1, 512, 2, 2] 512
Conv2d-16 [-1, 512, 2, 2] 262,144
InstanceNorm2d-17 [-1, 512, 2, 2] 1,024
PReLU-18 [-1, 512, 2, 2] 512
Upsample-19 [-1, 512, 4, 4] 0
Conv2d-20 [-1, 512, 4, 4] 2,359,296
InstanceNorm2d-21 [-1, 512, 4, 4] 1,024
PReLU-22 [-1, 512, 4, 4] 512
Conv2d-23 [-1, 512, 4, 4] 262,144
Upsample-24 [-1, 512, 4, 4] 0
Conv2d-25 [-1, 512, 4, 4] 262,144
InstanceNorm2d-26 [-1, 512, 4, 4] 1,024
_BlockBo...neckNd-27 [-1, 512, 4, 4] 0
_BlockResStkNd-28 [-1, 512, 4, 4] 0
InstanceNorm2d-29 [-1, 512, 4, 4] 1,024
PReLU-30 [-1, 512, 4, 4] 512
Conv2d-31 [-1, 512, 4, 4] 262,144
InstanceNorm2d-32 [-1, 512, 4, 4] 1,024
PReLU-33 [-1, 512, 4, 4] 512
Conv2d-34 [-1, 512, 4, 4] 2,359,296
InstanceNorm2d-35 [-1, 512, 4, 4] 1,024
PReLU-36 [-1, 512, 4, 4] 512
Conv2d-37 [-1, 256, 4, 4] 131,072
Conv2d-38 [-1, 256, 4, 4] 131,072
InstanceNorm2d-39 [-1, 256, 4, 4] 512
_BlockBo...neckNd-40 [-1, 256, 4, 4] 0
InstanceNorm2d-41 [-1, 256, 4, 4] 512
PReLU-42 [-1, 256, 4, 4] 256
Conv2d-43 [-1, 256, 4, 4] 65,536
InstanceNorm2d-44 [-1, 256, 4, 4] 512
PReLU-45 [-1, 256, 4, 4] 256
Upsample-46 [-1, 256, 8, 8] 0
Conv2d-47 [-1, 256, 8, 8] 589,824
InstanceNorm2d-48 [-1, 256, 8, 8] 512
PReLU-49 [-1, 256, 8, 8] 256
Conv2d-50 [-1, 256, 8, 8] 65,536
Upsample-51 [-1, 256, 8, 8] 0
Conv2d-52 [-1, 256, 8, 8] 65,536
InstanceNorm2d-53 [-1, 256, 8, 8] 512
_BlockBo...neckNd-54 [-1, 256, 8, 8] 0
_BlockResStkNd-55 [-1, 256, 8, 8] 0
InstanceNorm2d-56 [-1, 256, 8, 8] 512
PReLU-57 [-1, 256, 8, 8] 256
Conv2d-58 [-1, 256, 8, 8] 65,536
InstanceNorm2d-59 [-1, 256, 8, 8] 512
PReLU-60 [-1, 256, 8, 8] 256
Conv2d-61 [-1, 256, 8, 8] 589,824
InstanceNorm2d-62 [-1, 256, 8, 8] 512
PReLU-63 [-1, 256, 8, 8] 256
Conv2d-64 [-1, 128, 8, 8] 32,768
Conv2d-65 [-1, 128, 8, 8] 32,768
InstanceNorm2d-66 [-1, 128, 8, 8] 256
_BlockBo...neckNd-67 [-1, 128, 8, 8] 0
InstanceNorm2d-68 [-1, 128, 8, 8] 256
PReLU-69 [-1, 128, 8, 8] 128
Conv2d-70 [-1, 128, 8, 8] 16,384
InstanceNorm2d-71 [-1, 128, 8, 8] 256
PReLU-72 [-1, 128, 8, 8] 128
Upsample-73 [-1, 128, 16, 16] 0
Conv2d-74 [-1, 128, 16, 16] 147,456
InstanceNorm2d-75 [-1, 128, 16, 16] 256
PReLU-76 [-1, 128, 16, 16] 128
Conv2d-77 [-1, 128, 16, 16] 16,384
Upsample-78 [-1, 128, 16, 16] 0
Conv2d-79 [-1, 128, 16, 16] 16,384
InstanceNorm2d-80 [-1, 128, 16, 16] 256
_BlockBo...neckNd-81 [-1, 128, 16, 16] 0
_BlockResStkNd-82 [-1, 128, 16, 16] 0
InstanceNorm2d-83 [-1, 128, 16, 16] 256
PReLU-84 [-1, 128, 16, 16] 128
Conv2d-85 [-1, 128, 16, 16] 16,384
InstanceNorm2d-86 [-1, 128, 16, 16] 256
PReLU-87 [-1, 128, 16, 16] 128
Conv2d-88 [-1, 128, 16, 16] 147,456
InstanceNorm2d-89 [-1, 128, 16, 16] 256
PReLU-90 [-1, 128, 16, 16] 128
Conv2d-91 [-1, 64, 16, 16] 8,192
Conv2d-92 [-1, 64, 16, 16] 8,192
InstanceNorm2d-93 [-1, 64, 16, 16] 128
_BlockBo...neckNd-94 [-1, 64, 16, 16] 0
InstanceNorm2d-95 [-1, 64, 16, 16] 128
PReLU-96 [-1, 64, 16, 16] 64
Conv2d-97 [-1, 64, 16, 16] 4,096
InstanceNorm2d-98 [-1, 64, 16, 16] 128
PReLU-99 [-1, 64, 16, 16] 64
Upsample-100 [-1, 64, 32, 32] 0
Conv2d-101 [-1, 64, 32, 32] 36,864
InstanceNorm2d-102 [-1, 64, 32, 32] 128
PReLU-103 [-1, 64, 32, 32] 64
Conv2d-104 [-1, 64, 32, 32] 4,096
Upsample-105 [-1, 64, 32, 32] 0
Conv2d-106 [-1, 64, 32, 32] 4,096
InstanceNorm2d-107 [-1, 64, 32, 32] 128
_BlockBo...eckNd-108 [-1, 64, 32, 32] 0
_BlockResStkNd-109 [-1, 64, 32, 32] 0
InstanceNorm2d-110 [-1, 64, 32, 32] 128
PReLU-111 [-1, 64, 32, 32] 64
Conv2d-112 [-1, 64, 32, 32] 4,096
InstanceNorm2d-113 [-1, 64, 32, 32] 128
PReLU-114 [-1, 64, 32, 32] 64
Conv2d-115 [-1, 64, 32, 32] 36,864
InstanceNorm2d-116 [-1, 64, 32, 32] 128
PReLU-117 [-1, 64, 32, 32] 64
Conv2d-118 [-1, 64, 32, 32] 4,096
_BlockBo...eckNd-119 [-1, 64, 32, 32] 0
InstanceNorm2d-120 [-1, 64, 32, 32] 128
PReLU-121 [-1, 64, 32, 32] 64
Conv2d-122 [-1, 64, 32, 32] 4,096
InstanceNorm2d-123 [-1, 64, 32, 32] 128
PReLU-124 [-1, 64, 32, 32] 64
Upsample-125 [-1, 64, 64, 64] 0
Conv2d-126 [-1, 64, 64, 64] 36,864
InstanceNorm2d-127 [-1, 64, 64, 64] 128
PReLU-128 [-1, 64, 64, 64] 64
Conv2d-129 [-1, 64, 64, 64] 4,096
Upsample-130 [-1, 64, 64, 64] 0
Conv2d-131 [-1, 64, 64, 64] 4,096
InstanceNorm2d-132 [-1, 64, 64, 64] 128
_BlockBo...eckNd-133 [-1, 64, 64, 64] 0
_BlockResStkNd-134 [-1, 64, 64, 64] 0
Conv2d-135 [-1, 3, 64, 63] 4,803
DecoderNet2d-136 [-1, 3, 64, 63] 0
================================================================
Total params: 29,064,195
Trainable params: 29,064,195
Non-trainable params: 0
----------------------------------------------------------------
Input size (MB): 0.02
Forward/backward pass size (MB): 42.95
Params size (MB): 110.87
Estimated Total Size (MB): 153.84
----------------------------------------------------------------
Last update: March 14, 2021