Artificial Intelligence Model Interpreting Tools: SHAP, LIME, and Anchor Implementation in CNN Model for Hand Gestures Recognition

Explainable AI (XAI) are the tools and frameworks of artificial intelligence applications that make it easier to trust the results and outcomes produced by machine learning algorithms. Additionally, XAI helps with debugging, enhancing model performance, and describing the behavior of models to others. This paper presents an innovative approach for hand-gesture detection using an Explainable AI Convolutional Neural Network (XAI-CNN) and SHAP (Shapley Additive Explanations) values, LIME (Local Interpretable Model-agnostic Explanations), and Anchor as Explainable AI tools. The XAI-CNN model is specifically designed for ten different classes of hand-gesture accurate recognition, including palm moved, C, ok, I, fist, index, palm, thumb, down, and fist moved symbols. The proposed XAI-CNN architecture, built upon the previous CNN model, demonstrates an impressive accuracy of 99.98%. Furthermore, the SHAP (XAI tools) values, LIME, and Anchor integration enable the interpretation and visualization of the model's decision-making process separately, enhancing the transparency and trustworthiness in the hand-gesture recognition process. This research contributes to the robustness and interpretable AI systems for hand-gesture recognition, empowering users with accurate and understandable AI technology.