YawDD

In this paper, we present two video datasets of drivers with various facial characteristics, to be used for designing and testing algorithms and models for yawning detection. For collecting these videos, male and female candidates were asked to sit in the driver's seat of a car. The videos are taken in real and varying illumination conditions. In the first dataset, the camera is installed under the front mirror of the car. Each participant has three or four videos and each video contains different mouth conditions such as normal, talking/singing, and yawning. In the second dataset, the camera is installed on the dash in front of the driver, and each participant has one video with the above-mentioned different mouth conditions all in the same video. The car was parked for both datasets to keep the environment safe for the participants. As a benchmark, we also present the results of our own yawning detection method, and show that we can achieve a much higher accuracy in the scenario with the camera installed on the dash in front of the driver.