Sleep Deprivation Detection for Real-Time Driver Monitoring using Deep Learning

Miguel García-García (1) , Alice Caplier (1) , Michèle Rombaut (1)

1 GIPSA-AGPIG - GIPSA - Architecture, Géométrie, Perception, Images, Gestes

We propose a non-invasive method to detect sleep deprivation by evaluating a short video sequence of a subject. Computer Vision techniques are used to crop the face from every frame and classify it (within a Deep Learning framework) into two classes: " rested " or " sleep deprived ". The system has been trained on a database of subjects recorded under severe sleep deprivation conditions. A prototype has been implemented in a low-cost Android device proving its viability for real-time driver monitoring applications. Tests on real world data have been carried out and show encouraging performances but also reveal the need of larger datasets for training.

Mots clés: mobilenet; road safety; driver drowsiness; sleep deprivation

Domaines: Traitement du signal et de l'image [eess.SP]

Miguel García-García, Alice Caplier, Michèle Rombaut. Sleep Deprivation Detection for Real-Time Driver Monitoring using Deep Learning. ICIAR 2018 - 15th International Conference on Image Analysis and Recognition, Jun 2018, Povoa de Varzim, Portugal. ⟨hal-01837080⟩