Feature Configuration

Feature Configuration

The feature configuration allows you to chose what feature to use, and to adjust their parameters to your needs. Follow the steps below to create a YAML configuration file

1. YAML Configuration Structure

The configuration file should have the following structure

report_sections:
  - name: Section Name
    features:
      - FeatureName1
      - FeatureName2
      - FeatureName3

• report_sections: A list of sections that will appear in the final report. Each section consists of a name and a list of features to be included.
• name: The name of the section.
• features: The list of feature names to be included in the section.

2. Available Features

Please refer to the default configuration files to explore the available features and their names.

• Object Detection features - located in src/data_gradients/config/detection.yaml
• Semantic Segmentation features - located in src/data_gradients/config/segmentation.yaml

For a more in-depth explanation of each feature, please check out this page.

3. Feature Customization

Each feature can be customized by providing additional arguments in the configuration. Here's an example of customizing the DetectionSampleVisualization feature:

report_sections:
  - name: Object Detection Features
    features:
      - DetectionSampleVisualization:
          n_rows: 6
          n_cols: 2
          stack_splits_vertically: true
      - ... # Add any other features here

• DetectionSampleVisualization: The feature name.
• n_rows, n_cols: The number of rows and columns to use for displaying samples.
• stack_splits_vertically: Whether to show train/test samples vertically or side by side.

4. Using the Configuration

To use the configuration, provide the path of your YAML file to the relevant analysis manager. For example, for object detection analysis:

from data_gradients.managers.detection_manager import DetectionAnalysisManager

train_data = ...
val_data = ...
class_names = ...

analyzer = DetectionAnalysisManager(
    report_title="Testing Data-Gradients Object Detection",
    train_data=train_data,
    val_data=val_data,
    class_names=class_names,
    config_path="path/to/custom-detection-config.yaml"  # Add this parameter to the manager initialization
)

analyzer.run()

By following these steps, you can easily customize the features included in your analysis report and fine-tune their parameters according to your requirements.