crumpets.workers package

class crumpets.workers.ClassificationWorker(image, label, image_params=None, image_rng=None, **kwargs)[source]

Bases: crumpets.workers.ImageWorker

Worker for processing (Image, Label)-pairs for classification.

Parameters

  • image – tuple of image information (shape, dtype, fill_value); fill_value is optional, defaults to 0

  • label – tuple of label information (shape, dtype, fill_value); fill_value is optional, defaults to 0

  • image_params – dict of fixed image parameters; overwrites random augmentation values

  • image_rng – RNG object used for image augmentation, see RNG and randomize_args()

prepare(sample, batch, buffers)[source]

Implement this method to define the behavior of the BufferWorker subclass. Results must be written to buffers and/or batch object.

Parameters

  • sample – individual sample object to process

  • batch – the object the sample belongs to; append values to lists as necessary

  • buffers – output buffers to use for this sample

class crumpets.workers.FCNWorker(image, target_image, image_params=None, target_image_params=None, image_rng=None, **kwargs)[source]

Bases: crumpets.workers.ImageWorker

Worker for fully convolutional networks (FCN). Produces image-target_image-pairs.

Parameters

  • image – tuple of image information (shape, dtype, fill_value); fill_value is optional, defaults to 0

  • target_image – tuple of target image information (shape, dtype, fill_value); fill_value is optional, defaults to 0

  • image_params – dict of fixed image parameters; overwrites random augmentation values

  • target_image_params – dict of fixed target image parameters; overwrites random augmentation values

  • image_rng – RNG object used for image augmentation, see RNG and randomize_args()

prepare(sample, batch, buffers)[source]

Implement this method to define the behavior of the BufferWorker subclass. Results must be written to buffers and/or batch object.

Parameters

  • sample – individual sample object to process

  • batch – the object the sample belongs to; append values to lists as necessary

  • buffers – output buffers to use for this sample

class crumpets.workers.ImageWorker(image, image_params=None, image_rng=None, **kwargs)[source]

Bases: crumpets.broker.BufferWorker

Worker for processing images of any kind.

Parameters

  • image – tuple of image information (shape, dtype, fill_value); fill_value is optional, defaults to 0

  • image_params – dict of fixed image parameters; overwrites random augmentation values

  • image_rng – RNG object used for image augmentation, see RNG and randomize_args()

  • gpu_augmentation – disables augmentations for which gpu versions are available (randomizer)

prepare(sample, batch, buffers)[source]

Implement this method to define the behavior of the BufferWorker subclass. Results must be written to buffers and/or batch object.

Parameters

  • sample – individual sample object to process

  • batch – the object the sample belongs to; append values to lists as necessary

  • buffers – output buffers to use for this sample

prepare_image(im, buffers, params, key)[source]