Deep Learning for Non-Invasive Intoxication Detection: Ocular Analysis of Drivers Using CNNs and Quanvolutional Models

Current intoxication methods, though effective, often rely on intrusive procedures that require trained personnel or direct interaction with the individual, limiting their practicality for real-time applications [1], [2]. This study analyzes non-invasive approaches utilizing two Convolutional Neural Networks (CNNs) and a Quanvolutional Neural Network (QNN) trained on eye images to classify sobriety by analyzing visual cues such as pupil dilation, eye redness, and eyelid position. All models were trained and evaluated on a dataset of 1500 labeled images, achieving accuracies ranging from 68.65% to 99.74%.