How-To Guides

This section provides task-specific guides demonstrating how to use NRTK to apply perturbations, assess model robustness, and visualize system behavior. These examples are organized into general-purpose applications and those specifically integrated with Modular AI Trustworthy Engineering (MAITE) workflows.

Each guide links to a Jupyter notebook in the docs/examples/ directory of the repository.

General NRTK Examples

  • Visualizing Optical Transfer Functions

Explore and visualize different Optical Transfer Functions (OTFs) to understand their impact on image quality. View notebook.

  • Applying Image Perturbations

Use various image perturbation methods to simulate real-world distortions and evaluate model robustness. View notebook.

  • Applying Albumentations Perturbations via NRTK

Explore and visualize Albumentations perturbations in an NRTK context. View notebook.

  • Evaluating Models with COCO Scoring

Use COCO scoring to assess object detection model performance across perturbed inputs. View notebook.

Testing & Evaluation Guides with MAITE

The following notebooks showcase how NRTK perturbations can be applied to simulate key operational risks within a T&E workflow. Each notebook illustrates potential impact on model performance, utilizing MAITE as an evaluation harness.

  • Combining Perturbations with Saliency Maps

Integrate NRTK perturbations with saliency map generation to visualize how image changes affect model interpretation. View notebook.

  • Demonstrating Extreme Illumination Perturbations

Simulate brightness changes and evaluate model responses under lighting variability. View notebook.

  • Demonstrating Visual Focus Perturbations

Apply blur and focus distortions to test performance degradation from defocus. View notebook.

  • Demonstrating Fog or Haze Perturbations

Evaluate model robustness under haze-like visibility conditions using synthetic perturbations. View notebook.

  • Demonstrating Lens Flare Perturbations

Simulate a lens flare effect on an image and analyze its average and worst case effects on model precision. View notebook.

  • Demonstrating Resolution and Noise Transformations

Explore how camera-specific transformations affect model inputs and predictions. View notebook.

  • Demonstrating Random Translation Perturbations

Introduce pixel-level translations and observe model sensitivity to spatial shifts. View notebook.

  • Demonstrating Atmospheric Turbulence Perturbations

Simulate atmospheric distortion effects and assess their impact on image quality and model inference. View notebook.