Abstract: With increase in the population, accident rates are also increasing rapidly, the main reason is drowsiness of the driver. Such lethal incidents can be prevented if the driver is warned in time. To implement this technology, we proposes a smart alarm system to detect the drowsiness of driver using facial expressions and eye movements. Here open computer vision is used to detect driver’s eye movements for a long time.
We propose an approach based on Convolutional Neural Networks (CNN) that describes the object detection problem as sleepy detection. Based on the drivers' real-time video feed, it can detect and identify whether the eyes are open or closed. The technology used in this object detection challenge is the cellular CNN architecture with a single-shot multi-box detector. A different algorithm is used based on the output produced by the SSD_MobileNet_v1 architecture. A dataset of approximately 4,500 photographs of yawning, non-yawning, eyes-open, and eyes-closed subject faces was labelled to train the SSD_MobileNet_v1 network. The trained model is tested on about 600 randomly selected photos. The suggested strategy will guarantee improved computing efficiency and accuracy.
Keywords: single shot multi-box detector, Deep learning, Smart alarm, eye tracking, drowsiness detection.
|
DOI:
10.17148/IJARCCE.2023.12566
