Xaitk-Saliency: An Open Source Explainable AI Toolkit for Saliency

Advances in artificial intelligence (AI) using techniques such as deep learning have fueled the recent progress fields computer vision. However, these algorithms are still often viewed "black boxes", which cannot easily explain how they arrived at their final output decisions. Saliency maps one commonly used form of explainable AI (XAI), indicate input features an algorithm paid attention to during its decision process. Here, we introduce open source xaitk-saliency package, XAI framework and toolkit for saliency. We demonstrate modular flexible nature by highlighting two example use cases saliency maps: (1) object detection model comparison (2) doppelganger person re-identification. also show package can be paired with visualization tools support interactive exploration maps. Our results suggest that may play a critical role verification validation models, ensuring trusted deployment. The code is publicly available at: https://github.com/xaitk/xaitk-saliency.