vis4d.data.transforms.autoaugment

A wrap for timm transforms.

Classes

AugMix(*[, in_keys, out_keys, sensors, ...])	Apply Timm's AugMix to a image tensor.
AutoAugOriginal([magnitude_std])	Apply Timm's AutoAugment (policy=original) to a image array.
AutoAugV0([magnitude_std])	Apply Timm's AutoAugment (policy=v0) to a image array.
RandAug(*[, in_keys, out_keys, sensors, ...])	Apply Timm's RandomAugment to a image tensor.

class AutoAugV0(magnitude_std=0.5)

Apply Timm’s AutoAugment (policy=v0) to a image array.

__init__(magnitude_std=0.5)

Create an instance of AutoAug.

Parameters:

magnitude_std (float, optional) – Standard deviation of the magnitude for random autoaugment. Defaults to 0.5.

class AutoAugOriginal(magnitude_std=0.5)

Apply Timm’s AutoAugment (policy=original) to a image array.

__init__(magnitude_std=0.5)

Create an instance of AutoAug.

Parameters:

magnitude_std (float, optional) – Standard deviation of the magnitude for random autoaugment. Defaults to 0.5.

class RandAug(*, in_keys=['images'], out_keys=['images'], sensors=None, same_on_batch=True, **kwargs)

Apply Timm’s RandomAugment to a image tensor.

__call__(images)

Execute the transform.

Return type:

list[ndarray[Any, dtype[uint8]]]

class AugMix(*, in_keys=['images'], out_keys=['images'], sensors=None, same_on_batch=True, **kwargs)

Apply Timm’s AugMix to a image tensor.

__call__(images)

Execute the transform.

Return type:

list[ndarray[Any, dtype[uint8]]]