ShapBPT in Perspective: A Consolidated Review and an eXplainable Anomaly Detection Case Study
Date:
This talk presents ShapBPT in Perspective, a consolidated review and practical case study of ShapBPT for eXplainable Anomaly Detection (XAD).
The work connects hierarchical Shapley-based explanations with real-world anomaly detection systems, showing how structured image feature attributions can support the interpretation of black-box anomaly detection models.
🔗 Resources
📌 Talk Summary
- Presented ShapBPT as a data-aware hierarchical explanation method.
- Discussed its role in Explainable Computer Vision.
- Applied ShapBPT to visual anomaly detection.
- Showed how explanations can help interpret black-box anomaly detection systems.
⚙️ Method Overview
ShapBPT explains anomaly detection decisions by assigning attribution scores to image regions. Instead of using fixed geometric partitions, it relies on a Binary Partition Tree (BPT) to follow the intrinsic structure of the image.
🔍 Explainable Anomaly Detection Workflow
🖼️ Sample Output
Keywords
ShapBPT · Explainable Anomaly Detection · XAI · Shapley Values · Binary Partition Trees · Computer Vision · ICPE 2026 · QualITA Workshop