Transforms

In this section are defined the classes and functions to transform an image or video.

Note

Whilst some of these augmentations accept batched data, it will only sample random parameters once which might be detrimental to an effective learning.

Note

Some transforms rely on Torchvision functional.

Wrappers and Bases

List of wrappers of the data augmentations.

Wrapper(transforms[, inplace])	Wrap transforms to handle tensor data.
ImageWrapper(transforms[, inplace])	Wrap transforms to handle image data.
VideoWrapper(transforms[, inplace, video_format])	Wrap transforms to handle video data.

List of bases for the data augmentations and wrappers.

VideoBase(video_format)

Abstract class to make a base class for all video transforms.

Transforms

List of the data augmentations.

Div255([inplace])	Divide a tensor by 255.
MixUp(alpha[, inplace])	Mix input tensor with linear interpolation drawn according a Beta law.
Mul255([inplace])	Multiply a tensor by 255.
Normalize(mean, std[, cast_dtype, inplace, ...])	Normalize a tensor image with mean and standard deviation.
RandomApply(transforms[, p])	Apply randomly a list of transformations with a given probability.
RandomColorJitter([brightness, contrast, ...])	Randomly change the brightness, contrast, saturation and hue to images.
RandomGaussianBlur(kernel_size[, sigma, p, ...])	Blurs image with randomly chosen Gaussian blur.
RandomSolarize(threshold[, p, value_check])	Solarize the image randomly with a given probability by inverting all pixel values above a threshold.
RandomTransform(p)	Abstract class to make a base class for all random transforms.
VideoNormalize([mean, std, cast_dtype, ...])	Normalize a tensor video with mean and standard deviation.

Functional

List of the functions to make the data augmentations.

adjust_hue(img, hue_factor)	Adjust hue of an image.
div_255(tensor[, inplace])	Divide the given tensor by 255.
gaussian_blur(img, kernel_size[, sigma, ...])	Performs Gaussian blurring on the image by given kernel.
mixup(tensor_1, tensor_2, lam[, inplace])	Mix two tensors with linear interpolation.
mul_255(tensor[, inplace])	Multiply the given tensor by 255.
normalize(tensor, mean, std[, cast_dtype, ...])	Normalize a tensor image with mean and standard deviation.
solarize(img, threshold[, value_check])	Solarize an RGB/grayscale image by inverting all pixel values above a threshold.