Shared Attributes

Description

The seamless interaction between research modules is enabled by shared attributes defined in the ResearchAttributes class. This class contains attributes that are used across the research modules through inheritance.

Class

class ResearchAttributes(label_type=None, class_names=None)

Store attributes shared between modules in the research package.

__init__(label_type=None, class_names=None)

Initialize the ResearchAttributes class.

Parameters:

• label_type (str or None) – The type of labels used: ‘binary’, ‘multi_class’, ‘multi_label’, ‘multi_label_multi_class’, ‘object_detection’. If None, label_manager is set to None.

• class_names (list or None) – The list of class names.

datasets_container

Dictionary containing datasets. When creating new datasets, ‘complete_dataset’ is added; when split, ‘train_dataset’, ‘val_dataset’, and ‘test_dataset’ are added.

Type:

dict

label_manager

LabelManager instance for handling labels.

Type:

LabelManager

outputs_container

Dictionary containing outputs in the form of tuples (y_true, y_pred). Keys follow the pattern: dataset name with ‘dataset’ replaced by ‘outputs’.

Type:

dict

model

The Keras model instance.

Type:

tf.keras.Model

training_history

Dictionary tracking the training history of the model after fitting.

Type:

dict

evaluation_metrics

Tracked evaluation metrics of the model after evaluating. Format: {Set_Name: {Metric: Value}}.

Type:

dict

figures

Dictionary containing tracked figures. Format: {figure_name: figure}.

Type:

dict

synchronize_research_attributes(research_attributes)

Synchronize research attributes with another ResearchAttributes instance.

Parameters:

research_attributes (ResearchAttributes) – The instance to synchronize with.

reset_research_attributes(except_datasets=False)

Reset research attributes while preserving the label manager.

Parameters:

except_datasets (bool) – If True, datasets are not reset. Defaults to False.