cerebras.modelzoo.data.vision.segmentation.transforms.noise_transforms.GaussianNoiseTransform#

class cerebras.modelzoo.data.vision.segmentation.transforms.noise_transforms.GaussianNoiseTransform(noise_variance=(0, 0.1), p_per_sample=1, p_per_channel=1, per_channel=False, data_key='data')[source]#

Bases: object

Adds additive Gaussian Noise

Parameters

  • noise_variance – variance is uniformly sampled from that range

  • p_per_sample –

  • p_per_channel (float) –

  • per_channel (bool) – if True, each channel will get its own variance sampled from noise_variance

  • data_key –

CAREFUL: This transform will modify the value range of your data!

Methods

previous

cerebras.modelzoo.data.vision.segmentation.transforms.noise_transforms

next

cerebras.modelzoo.data.vision.segmentation.transforms.resample_augmentations