From face images to a level of drowsiness

M. Quentin Massoz défendra publiquement sa thèse intitulé "Non-invasive, automatic, and real-time characterization of drowsiness based on eye closure dynamics »

Drowsiness is a complex physiological state associated with a difficulty of staying awake, a strong inclination towards falling asleep. When performing a task, drowsiness impairs the ability of an individual to make sound decisions and to complete said task at adequate performance. When the task is critical, drowsiness therefore becomes a danger that puts human lives at risk. In facts, drowsiness is a major cause of fatal accidents, in particular in the transportation sector where it is estimated to be responsible for 20–30% of them. There is thus a clear need for automatic, real-time drowsiness characterization systems that aim at preventing such accidents by issuing timely drowsiness warnings to vehicle operators.

This thesis is about drowsiness characterization systems that operate non-invasively, automatically, and in real-time from a video stream of face images, with a focus on the analysis of eye closure dynamics. In addition to providing a comprehensive discussion about the development of such systems, we present three novel systems: a baseline system, a multi-timescale system, and a parametric system. We thoroughly evaluate the performance of our machine-learning-based systems, and provide key insights on how they operate/make their decisions.

Promoteur : Marc Van Droogenbroeck

Informations pratiques

La défense (en anglais) est accessible à tous et aura lieu le mercredi 24 avril 2019 à 9h00, dans le local R7 de l'Institut Montefiore, Bâtiment B28, au Sart Tilman (accès et plan).