Filters

Implements filters used by models.

class anomalib.models.components.filters.GaussianBlur2d(sigma, channels=1, kernel_size=None, normalize=True, border_type='reflect', padding='same')

Bases: Module

Compute GaussianBlur in 2d.

Makes use of kornia functions, but most notably the kernel is not computed during the forward pass, and does not depend on the input size. As a caveat, the number of channels that are expected have to be provided during initialization.

forward(input_tensor)

Blur the input with the computed Gaussian.

Parameters:

input_tensor (torch.Tensor) – Input tensor to be blurred.

Returns:

Blurred output tensor.

Return type:

Tensor