Eyes wide open? eyelid location and eye aperture estimation for pervasive eye tracking in real-world scenarios
Wolfgang Fuhl, Thiago Santini, David Geisler, Thomas Kübler, Wolfgang Rosenstiel, Enkelejda Kasneci · 2016
Eyelid identification and aperture estimation provide key data that can be used to infer valuable information about a subject's mental state (e.g., vigilance, fatigue, and drowsiness) as well as validate or reduce the search space of other eye features. In this paper, we consider these tasks from the perspective of pervasive eye tracking, taking into account the multiple challenges and constraints that arise from this scenario. A novel method for eyelid identification and aperture estimation is proposed and evaluated against challenging data from an eye-tracking experiment conducted in driving scenarios in the wild. The proposed method outperforms an state-of-the-art approach by up to 40 percentage points and runs in real-time on state-of-the-art eye tracking systems. The method implementation and the realistic dataset are provided openly at www.ti.uni-tuebingen.de/perception.