Catalog of Pre-Trained Models For Driver Drowsiness Detection

Every year, thousands of lives are lost on the roads, posing significant challenges to public health and road safety. Recent data from the Federal Highway Police (PRF) indicate that more than 25% of fatalities on federal highways are associated with critical behaviors, such as driver drowsiness, delayed reactions, or complete lack of response. To address this issue, advanced computer vision techniques have emerged as effective tools for the early detection of drowsiness in drivers. In particular, the use of pre-trained models has gained prominence due to their ability to leverage extensive training on large datasets, which enables the precise extraction of facial features and state classification (for example, distinguishing between open or closed eyes and detecting yawns). Despite the wide availability of pre-trained models, novice researchers frequently encounter difficulties navigating the myriad options, comparing their performance, and determining which model best meets their specific needs. This article presents a systematic catalog of pre-trained models, synthesizing their key characteristics and providing guidance on where to access them. By organizing these resources, the catalog aims to simplify model selection and support the development of effective drowsiness monitoring systems that can help prevent accidents and enhance road safety.