Conditions

Conditions are useful for any custom pre- and post-processing that must be done on batches of data. Module trainer maintains two separate condition lists that are executed before/after the network forward pass.

An example of a condition could be an Assert that needs to be performed before data is processed. A more advanced example of a condition could be code that modifies the network based on input or output

class pywick.conditions.Condition[source]

Default class from which all other Condition implementations inherit.

reset()[source]
class pywick.conditions.SegmentationInputAsserts[source]

Executes segmentation-specific asserts before executing forward pass on inputs

reset()[source]
class pywick.conditions.SegmentationOutputAsserts(num_classes)[source]

Executes segmentation-specific asserts after executing forward pass on inputs

reset()[source]