In this paper, we define the fatigue detection problem in real-world environments and propose a heterogeneous multi-source fatigue detection framework that utilizes various sensors. Unlike existing methods that rely on expensive sensors and controlled environments, our study aims to enable practical fatigue monitoring even in limited sensor environments. Through experiments using real-world field-deployed sensor setups and public datasets, we demonstrate the practicality, robustness, and improved generalization performance of the proposed framework. This is an important step toward practical applications of fatigue detection in real-world environments, such as aviation, mining, and long-distance transportation.
