CRNN-Based Eye Behavior Analysis for Drowsiness Detection

Deep learning plays a critical role in enhancing road safety through various applications and technologies. Deep learning is at the core of Advanced Driver Assistance Systems, which provide real-time feedback and warnings to drivers. Drowsiness detection is a critical aspect of ensuring road safety and preventing accidents, especially in long-haul driving and transportation sectors. Deep learning models can analyze driver behavior, including eye movement and facial expressions, to detect signs of drowsiness. Drowsiness detection systems can then issue warnings suggesting rest breaks to prevent accidents caused by drowsy driving. This research explores the development and evaluation of a Convolutional Recurrent Neural Network architecture for drowsiness detection from eye images. The CRNN-based design utilizes a synergistic combination of CNN and LSTM layers, offering a comprehensive solution for drowsiness detection that considers both spatial and temporal aspects of eye behavior. The proposed model leverages the Eye Aspect Ratio and Eye Closure Ratio as key features for accurate detection. The proposed CRNN-based model is evaluated against traditional machine learning and deep learning models. Experimental results reveal that the CRNN-based model achieved an impressive accuracy of 93.2%, outperforming compared models and showcasing its potential for real-world applications.