class

CircularPad2d

extendsModule

CircularPad2d(padding: int | tuple[int, int, int, int])

source

Pad a 4-D tensor (N, C, H, W) with wrap-around in both spatial axes.

Circular padding in 2-D treats both the height and width axes as periodic: pixels from the right column wrap to the left, and pixels from the bottom row wrap to the top.

\text{output}[n, c, i, j] = x\!\left[n, c,\; (i - p_{\text{top}}) \bmod H,\; (j - p_{\text{left}}) \bmod W \right]

Parameters

paddingint or tuple[int, int, int, int]

(left, right, top, bottom) circular padding sizes. A single int pads equally on all four sides.

Attributes

paddingtuple[int, int, int, int]

Normalised (left, right, top, bottom) padding.

Notes

Input: $(N, C, H, W)$ .
Output: $(N, C, H + p_{\text{top}} + p_{\text{bottom}}, W + p_{\text{left}} + p_{\text{right}})$ .

Useful for processing panoramic (360°) images where left and right edges are physically adjacent, or for convolutional processing of spherical data projected to a 2-D grid.
Combining circular padding with a stride-2 convolution on a toroidal feature map maintains the topology of the feature space.

Examples

**Seamless tiling convolution for a panoramic image:**
>>> import lucid
>>> import lucid.nn as nn
>>>
>>> pad = nn.CircularPad2d(padding=(2, 2, 0, 0))  # horizontal wrap only
>>> x = lucid.zeros(1, 3, 256, 512)
>>> pad(x).shape
(1, 3, 256, 516)
**Uniform circular border:**
>>> pad = nn.CircularPad2d(padding=1)
>>> x = lucid.zeros(4, 16, 32, 32)
>>> pad(x).shape
(4, 16, 34, 34)

Methods (3)

dunder

init

→None

__init__(padding: int | tuple[int, int, int, int])

source

Initialise the CircularPad2d module. See the class docstring for parameter semantics.

forward

→Tensor

forward(x: Tensor)

source

Pad the input tensor according to the configured padding.

Parameters

inputTensor

Input tensor of shape

(N, C, *)

Returns

Tensor

Padded tensor with spatial dimensions expanded by the configured padding amounts.

extra_repr

→str

extra_repr()

source

Return a string representation of the layer's configuration.