ForestMonkey: Toolkit for Reasoning with AI-based Defect Detection and Classification Models

ForestMonkey enables non-experts to understand AI defect detection decisions: The toolkit automatically generates charts and text explanations that help users understand why an AI model made specific predictions about defects, making AI reasoning accessible without requiring deep technical knowledge.

The toolkit identifies specific ways to improve AI models: By analyzing prediction patterns and defect characteristics, ForestMonkey suggests concrete improvements to AI systems, helping organizations move beyond simply accepting model outputs to actively enhancing their performance.

Explainability is built into the workflow, not added afterward: Rather than treating explanation as a separate step, ForestMonkey integrates reasoning directly into the model evaluation process by extracting features from both predictions and images, embedding transparency into how defect detection systems operate.

The approach is practical and scalable across different AI models: The research tested ForestMonkey on four different AI models with various datasets and measured performance, demonstrating that this reasoning approach can work across diverse defect detection systems rather than being limited to one specific model type.

Users can implement this without specialized AI expertise: Packaged as a Python tool that accepts simple inputs (image folders, labels, predictions), ForestMonkey lowers barriers for organizations to adopt explainable AI practices in quality control and defect detection applications.