transforms.RandomGrayscale¶

class lucid.transforms.RandomGrayscale(p: float = 0.1)¶

The RandomGrayscale class is a transformation that randomly converts an input tensor to grayscale with a given probability. This transformation is often used in data augmentation to increase robustness in training image-based neural networks.

Class Signature¶

class lucid.transforms.RandomGrayscale(p: float = 0.1)

Parameters¶

p (float, optional): The probability of converting the input tensor to grayscale. Must be a value between 0 and 1. Default is 0.1.

Attributes¶

p (float): The probability with which the input tensor will be converted to grayscale, stored as an internal attribute.

Forward Calculation¶

The conversion of an input tensor \(\mathbf{x}\) to grayscale is applied with a probability \(p\) as follows:

Generate a random number \(r\) from a uniform distribution over [0, 1].
If \(r < p\), convert the input tensor to grayscale using the following formula for each pixel (or channel) \((r, g, b)\):

\[y = 0.299r + 0.587g + 0.114b\]

where \(r\), \(g\), and \(b\) are the red, green, and blue channels, respectively. The resulting value \(y\) replaces the color channels with a single grayscale intensity.

If the tensor has multiple channels (e.g., an image batch), the grayscale transformation is applied independently to each image.

Examples¶

Example 1: Basic Usage

>>> import lucid.transforms as T
>>> input_tensor = lucid.Tensor([[[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],
                                  [[7.0, 8.0, 9.0], [10.0, 11.0, 12.0]],
                                  [[13.0, 14.0, 15.0], [16.0, 17.0, 18.0]]]])  # Shape: (1, 3, 2, 3)
>>> grayscale = T.RandomGrayscale(p=1.0)  # Always convert to grayscale
>>> output = grayscale(input_tensor)
>>> print(output)
Tensor(...)  # The exact output will be a grayscale version of the input

Example 2: Randomized Grayscale

>>> import lucid.transforms as T
>>> input_tensor = lucid.Tensor([[[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]],
                                  [[7.0, 8.0, 9.0], [10.0, 11.0, 12.0]],
                                  [[13.0, 14.0, 15.0], [16.0, 17.0, 18.0]]]])  # Shape: (1, 3, 2, 3)
>>> grayscale = T.RandomGrayscale(p=0.5)  # 50% chance to convert to grayscale
>>> output = grayscale(input_tensor)  # Random outcome

Note

This transformation is useful for data augmentation, especially in scenarios where robustness to color changes is essential.
Setting p=1.0 ensures that the input tensor is always converted to grayscale, while p=0.0 ensures that the input tensor is never converted.
The grayscale transformation maintains the shape of the tensor, but the number of channels may be reduced depending on the implementation.