Explainable Anomaly Detection

Code

This project studies how Explainable AI can improve trust in image anomaly detection systems.

Overview

Anomaly detection systems are often used in safety-critical domains, but their outputs can be difficult to interpret. This project combines VAE-GAN-based anomaly detection with explanation techniques to help users understand why a sample is considered anomalous.

Main Contributions

• Reconstruction-based anomaly detection using VAE-GAN models.
• Visual anomaly maps for highlighting abnormal regions.
• Trust-oriented analysis of anomaly detection outputs.
• Case-study-based evaluation of explainability in anomaly detection.

Applications

• Industrial inspection.
• Safety-critical monitoring.
• Visual quality control.
• Human-centered AI systems.

Technologies

Python · PyTorch · VAE-GAN · OpenCV · Explainable AI · Anomaly Detection