Bottleneck

class lib.model.networks.clip.Bottleneck(inplanes: int, planes: int, stride: int = 1, name: str = 'bottleneck')

Bases: object

A ResNet bottleneck block that performs a sequence of convolutions, batch normalization, and ReLU activation operations on an input tensor.

Parameters:

inplanes (int) – The number of input channels.
planes (int) – The number of output channels.
stride (int, optional) – The stride of the bottleneck block. Default: 1
name (str, optional) – The name of the bottleneck block. Default: “bottleneck”

Attributes Summary

The factor by which the number of input channels is expanded to get the number of output channels.

Methods Summary

Performs the forward pass for a Bottleneck block.

Attributes Documentation

expansion = 4

The factor by which the number of input channels is expanded to get the number of output channels.

Methods Documentation

__call__(inputs: KerasTensor) → KerasTensor

Performs the forward pass for a Bottleneck block.

All conv layers have stride 1. an avgpool is performed after the second convolution when stride > 1

Parameters:: inputs (keras.KerasTensor) – The input tensor to the Bottleneck block.
Returns:: The result of the forward pass through the Bottleneck block.
Return type:: keras.KerasTensor

__call__(inputs: KerasTensor) → KerasTensor

Performs the forward pass for a Bottleneck block.

All conv layers have stride 1. an avgpool is performed after the second convolution when stride > 1

Parameters:: inputs (keras.KerasTensor) – The input tensor to the Bottleneck block.
Returns:: The result of the forward pass through the Bottleneck block.
Return type:: keras.KerasTensor

expansion = 4

The factor by which the number of input channels is expanded to get the number of output channels.