baby.augmentation

A set of augmentation functions for on-the-fly preprocessing of images before input into the Neural Network. All augmentation functions are meant to work with unstructured arrays (in particular, elastic_deform, turn, and the shifts probably will return errors if used on structured arrays. They all take exactly two arrays as input, and will perform identitcal transformations on both arrays.

Classes

Augmenter([xy_out, probs, p_noop, substacks])	Methods
DownscalingAugmenter(*args[, pixdev, xy_scaled])	Methods
ScalingAugmenter(*args[, target_pixel_size, ...])	Augmenter with a target pixel size
SmoothedLabelAugmenter(sigmafunc[, ...])	Methods
SmoothingSigmaModel([a, b, c])	Model for picking a smoothing sigma for gaussian filtering