AutoNN.CNN.models.resnet.resnet

def  resnet(architecture:int=18,**kwargs)->ResNet

This function can be used to call different Resnet architectures. All resnet architectures are implemted in PyTorch

Resnet	num_residual_block
18	[2,2,2,2]
34	[3,4,6,3]
50	[3,4,6,3]
101	[3,4,23,3]
152	[3,8,36,3]
custom	any combination you like

Example:

To use ResNet50 with 10 output classes use:

from AutoNN.CNN.models.resnet import resnet

model = resnet(architecture=50,num_class=10)
print(model)

If you want to define your ResNet with custom number of blocks and then use:

model = resnet(architecture=-1,in_channels=3,
            num_residual_block=[3,4,6,3],
            num_class=10,
            block_type='normal')

Arguments :

• architecture = -1 means you can choose the #layers of your resnet

• inchannels = number of input channels

• num_residual_blocks = number of residual blocks in each layer, [3,4,6,3] means the model has 4 layers with

  • 1st layer containing 3 residual blocks,
  • 2nd layer --> 4 residual blocks
  • 3rd layer --> 6 residual blocks
  • 4th layer --> 3 residual blocks

• num_class = number of classes (in this case it's 10)

• block_type = this has two types

  i) 'normal' : has residual block with (3x3 conv, 3x3 conv) in that order

  ii) 'botleneck' : has limeisual block with (1x1 conv, 3x3 conv, 1x1 conv) in that order

Returns :

Resnet() with desires number of layers and residual blocks

Example : Custom Model

    model = resnet(-1,num_residual_blocks=[2,1],num_class = 4,block_type='bottleneck')

Output

ResNet(
(conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
(bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(relu): ReLU(inplace=True)
(maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)
(resnet): Sequential(
    (0): Bottleneck(
    (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): ReLU(inplace=True)
    (downSample): Sequential(
        (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
        (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    )
    (1): Bottleneck(
    (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)
    (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): ReLU(inplace=True)
    )
    (2): Bottleneck(
    (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)
    (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)
    (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    (relu): ReLU(inplace=True)
    (downSample): Sequential(
        (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)
        (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
    )
    )
    )
    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))
    (fc): Linear(in_features=512, out_features=4, bias=True)
        )