stable_pretraining.data.transforms

Functions

random_seed(seed)
set_seed(seeds)
to_image(input)	See ToImage for details.

Classes

AdditiveGaussian(sigma[, p])	Add Gaussian noise to input data.
CenterCrop(size[, source, target])	Crop the center of an image to the given size.
ColorJitter([brightness, contrast, ...])	Randomly change brightness, contrast, saturation, and hue of an image.
Compose(*args)	Compose multiple transforms together in sequence.
Conditional(transform, condition_key[, ...])	Apply transform conditionally based on a data dictionary key.
ControlledTransform(transform[, ...])	Face Landmarks dataset.
GaussianBlur(kernel_size[, sigma, p, ...])	Apply Gaussian blur to image with random sigma values.
MultiViewTransform(transforms)	Apply different transforms to different views of the same sample.
PILGaussianBlur([sigma, p, source, target])	PIL-based Gaussian blur transform with random sigma sampling.
RGB([source, target])	Convert image to RGB format.
RandomChannelPermutation([source, target])	Randomly permute the channels of an image.
RandomContiguousTemporalSampler(source, ...)	Randomly sample contiguous frames from a video sequence.
RandomCrop(size[, padding, pad_if_needed, ...])	Crop a random portion of image and resize it to given size.
RandomGrayscale([p, source, target])	Randomly convert image to grayscale with given probability.
RandomHorizontalFlip([p, source, target])	Horizontally flip the given image randomly with a given probability.
RandomResizedCrop(size[, scale, ratio, ...])	Crop a random portion of image and resize it to given size.
RandomRotation(degrees[, interpolation, ...])	Rotate image by random angle within specified degrees range.
RandomSolarize(threshold[, p, source, target])	Randomly solarize image by inverting pixel values above threshold.
Resize(size[, interpolation, max_size, ...])	Resize image to specified size.
ToImage([dtype, scale, mean, std, source, ...])	Convert input to image tensor with optional normalization.
Transform()	Base transform class extending torchvision v2.Transform with nested data handling.
UniformTemporalSubsample(num_samples[, ...])	nn.Module wrapper for pytorchvideo.transforms.functional.uniform_temporal_subsample.