HeatmapComparison

class HeatmapComparison

Check if the average image brightness (or bbox location if applicable) is similar between train and test set.

The check computes the average grayscale image per dataset (train and test) and compares the resulting images. Additionally, in case of an object detection task, the check will compare the average locations of the bounding boxes between the datasets.

Parameters

classes_to_displayOptional[List[float]], default: None

List of classes to display in bounding box heatmap, using the class names (strings). Applies only for object detection tasks. If None, all classes are displayed.

__init__(classes_to_display: Optional[List[str]] = None, **kwargs)

__new__(*args, **kwargs)

Methods

HeatmapComparison.add_condition(name, ...)	Add new condition function to the check.
HeatmapComparison.clean_conditions()	Remove all conditions from this check instance.
HeatmapComparison.compute(context)	Create the average images and display them.
HeatmapComparison.conditions_decision(result)	Run conditions on given result.
HeatmapComparison.config()	Return check configuration (conditions' configuration not yet supported).
HeatmapComparison.from_config(conf)	Return check object from a CheckConfig object.
HeatmapComparison.initialize_run(context)	Initialize run.
HeatmapComparison.metadata([with_doc_link])	Return check metadata.
HeatmapComparison.name()	Name of class in split camel case.
HeatmapComparison.params([show_defaults])	Return parameters to show when printing the check.
HeatmapComparison.plot_row_of_heatmaps(...)	Plot a row of heatmaps for train and test images.
HeatmapComparison.remove_condition(index)	Remove given condition by index.
HeatmapComparison.run(train_dataset, ...[, ...])	Run check.
HeatmapComparison.update(context, batch, ...)	Perform update on batch for train or test counters and histograms.

Examples

Heatmap Comparison

previous

train_test_validation

next

HeatmapComparison.add_condition