DFBRFeatureExtractor¶

tiatoolbox . tools . registration . wsi_registration . DFBRFeatureExtractor

class DFBRFeatureExtractor[source]¶

Feature extractor for Deep Feature based Registration (DFBR).

This class extracts features from three different layers of VGG16. These features are processed in DFBRegister class for registration of a pair of images.

Initialize DFBRFeatureExtractor.

Methods

`forward`	Forward pass for feature extraction.
`forward_hook`	Register a hook.

Attributes

training

forward(x)[source]¶

Forward pass for feature extraction.

Parameters:

x (torch.Tensor) – Batch of input images.
self (Module)

Returns:

A dictionary containing the multiscale features. The expected format is {layer_name: features}.

Return type:

dict

forward_hook(layer_name)[source]¶

Parameters:

layer_name (str) – User-defined name for a layer.
self (Module)

Returns:

Forward hook for feature extraction.

Return type:

hook (Callable)