CIAO Method - Enterprise Architecture

Overview

This is a modernized implementation of the CIAO (Contextual Importance Assessment via Obfuscation) method for explainable AI, restructured to follow enterprise-grade architectural patterns with Hydra configuration management and PyTorch Lightning integration.

Features

• Hydra-based configuration management
• PyTorch Lightning integration
• Factory pattern for components
• Support for medical imaging datasets (prostate/colorectal cancer)
• Modular architecture for easy extensibility

Structure

ciao/
├── components/         # High-level CIAO components
│   ├── segmentation/
│   ├── obfuscation/
│   ├── explainer/
│   └── factory.py
├── modeling/          # Core model implementations
├── data/             # Data loading and processing
├── utils/            # Utility functions
└── scripts/          # Data processing scripts

Installation

uv sync

Usage

ciao
# or
python -m ciao

Configuration

The system uses Hydra for configuration management. See configs/ directory for available configurations.