nntoolbox.vision.components.se module

class nntoolbox.vision.components.se.SEBlock(in_channels, reduction_ratio=16)

Bases: torch.nn.modules.module.Module

Implement squeeze (global information embedding) and excitation (adaptive recalibration) mechanism:

References:

Jie Hu, Li Shen, Samuel Albanie, Gang Sun, Enhua Wu. “Squeeze-and-Excitation Networks.” https://arxiv.org/pdf/1709.01507.pdf

forward(input)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

training: bool

class nntoolbox.vision.components.se.SEResidualBlock(in_channels, reduction_ratio=16)

Bases: torch.nn.modules.container.Sequential

training: bool

class nntoolbox.vision.components.se.SEResidualBlockPreActivation(in_channels, reduction_ratio=16, activation=<class 'torch.nn.modules.activation.ReLU'>, normalization=<class 'torch.nn.modules.batchnorm.BatchNorm2d'>)

Bases: nntoolbox.vision.components.res.ResNeXtBlock

training: bool

class nntoolbox.vision.components.se.SEResidualBlockPreActivationKer(in_channels, kernel, reduction_ratio=16, activation=<class 'torch.nn.modules.activation.ReLU'>, normalization=<class 'torch.nn.modules.batchnorm.BatchNorm2d'>)

Bases: nntoolbox.vision.components.res.ResNeXtBlock

training: bool