Samplewise Standardizer

tefla.da.standardizer.SamplewiseStandardizer (clip, channel_wise=False)

Args

  • clip: max/min allowed value in the output image e.g.: 6
  • channel_wise: perform standarization separately accross channels

Samplewise Standardizer

tefla.da.standardizer.SamplewiseStandardizerTF (clip, channel_wise=False)

Args

  • clip: max/min allowed value in the output image e.g.: 6
  • channel_wise: perform standarization separately accross channels

Aggregate Standardizer

tefla.da.standardizer.AggregateStandardizer (mean, std, u, ev, sigma=0.0, color_vec=None)

Creates a standardizer based on whole training dataset

Args

  • mean: 1-D array, aggregate mean array e.g.: mean is calculated for each color channel, R, G, B
  • std: 1-D array, aggregate standard deviation array e.g.: std is calculated for each color channel, R, G, B
  • u: 2-D array, eigenvector for the color channel variation
  • ev: 1-D array, eigenvalues
  • sigma: float, noise factor
  • color_vec: an optional color vector

Methods

augment_color (img, sigma=0.0, color_vec=None)

Args
  • img: input image
  • sigma: a float, noise factor
  • color_vec: an optional color vec

Aggregate Standardizer

tefla.da.standardizer.AggregateStandardizerTF (mean, std, u, ev, sigma=0.0, color_vec=None)

Creates a standardizer based on whole training dataset

Args

  • mean: 1-D array, aggregate mean array e.g.: mean is calculated for each color channel, R, G, B
  • std: 1-D array, aggregate standard deviation array e.g.: std is calculated for each color channel, R, G, B
  • u: 2-D array, eigenvector for the color channel variation
  • ev: 1-D array, eigenvalues
  • sigma: float, noise factor
  • color_vec: an optional color vector

Methods

augment_color (img, sigma=0.0, color_vec=None)

Args
  • img: input image
  • sigma: a float, noise factor
  • color_vec: an optional color vec