ConvBnRelu¶

class ConvBnRelu(num_input_channels, num_output_channels, kernel_size=(3, 3), strides=(1, 1), dilation_rate=(1, 1), activation='relu', *, use_bias=False, do_batchnorm=True)[source]¶

Performs Convolution, Batch Normalization and activation.

Parameters:

  • num_input_channels (int) – Number of channels in input.

  • num_output_channels (int) – Number of channels in output.

  • kernel_size (int) – Size of the kernel in the convolution layer.

  • strides (int) – Size of the stride in the convolution layer.

  • use_bias (bool) – Whether to use bias in the convolution layer.

  • dilation_rate (int) – Dilation rate in the convolution layer.

  • activation (str) – Name of the activation function to use.

  • do_batchnorm (bool) – Whether to do batch normalization after the convolution layer.

Returns:

a pytorch model.

Return type:

model (torch.nn.Module)

Initialize ConvBnRelu.

Methods

forward

Logic for using layers defined in init.

get_block

Function to acquire a convolutional block.

Attributes

training

forward(input_tensor)[source]¶

Logic for using layers defined in init.

This method defines how layers are used in forward operation.

Parameters:
Returns:

The inference output.

Return type:

output (torch.Tensor)

static get_block(in_channels, out_channels, kernel_size, strides, dilation_rate, activation, *, do_batchnorm, use_bias)[source]¶

Function to acquire a convolutional block.

Parameters:

  • in_channels (int) – Number of channels in input.

  • out_channels (int) – Number of channels in output.

  • kernel_size (int or tuple(int, int)) – Size of the kernel in the acquired convolution block.

  • strides (int) – Size of stride in the convolution layer.

  • use_bias (bool) – Whether to use bias in the convolution layer.

  • dilation_rate (int or tuple(int, int)) – Dilation rate for each convolution layer.

  • activation (str) – Name of the activation function to use.

  • do_batchnorm (bool) – Whether to do batch normalization after the convolution layer.

Returns:

a pytorch layer

Return type:

torch.nn.Sequential