Fatigue Detection

Abstract

The author uses two video datasets for the experiment. For the datasets, the participants are asked to sit in the driver's seat of a parked car. Then videos are captured for various mouth conditions such as resting, talking/singing, and yawning under different illumination conditions. The author keep exactly the same environmental parameters to collect videos for both datasets. The only change is the camera angle. For the first dataset, the camera is placed under the front mirror. In the second dataset, the camera is placed on the dash of the car. The author eventually describes that they achieved a higher degree of accuracy for yawning detection when the camera is placed on the dash board for the camera. [1] Martin et al. introduces the first large-scale dataset for recognizing driver activity. The data is collected in both manual and autonomous driving modes. A wide variety of information such as color, IR, depth and 3D body pose is collected from six views and categorized into 83 categories. The data is collected using five NIR-and the Kinect v2 cameras. The major challenges involves