2019/04/15 by Reza Ghoddoosian, Ghoddoosian, Reza, Marnim Galib +3 · 4 citations
Computer Science · Engineering · Psychology · #Artificial intelligence #Baseline (sea) #Benchmark (surveying) #Cartography #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Core (optical fiber) #FOS: Computer and information sciences #Fire Detection and Safety Systems #Geography #Machine learning #Pattern recognition (psychology) #Ranging #Sleep and Work-Related Fatigue #Telecommunications #cs.CV
paper · pdf · doi:10.48550/arxiv.1904.07312
Computer Vision and Pattern Recognition Workshops (CVPRW 2019)
arxiv created 2019/04/15 · openalex publication_date 2019/04/15 · arxiv updated 2019/04/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06
Drowsiness can put lives of many drivers and workers in danger. It is\nimportant to design practical and easy-to-deploy real-world systems to detect\nthe onset of drowsiness.In this paper, we address early drowsiness detection,\nwhich can provide early alerts and offer subjects ample time to react. We\npresent a large and public real-life dataset of 60 subjects, with video\nsegments labeled as alert, low vigilant, or drowsy. This dataset consists of\naround 30 hours of video, with contents ranging from subtle signs of drowsiness\nto more obvious ones. We also benchmark a temporal model for our dataset, which\nhas low computational and storage demands. The core of our proposed method is a\nHierarchical Multiscale Long Short-Term Memory (HM-LSTM) network, that is fed\nby detected blink features in sequence. Our experiments demonstrate the\nrelationship between the sequential blink features and drowsiness. In the\nexperimental results, our baseline method produces higher accuracy than human\njudgment.\n