Researchers from the Vietnam National University of Ho Chi Minh City have developed a convolutional neural network, a deep learning algorithm that is often used to analyze visual imagery, to determine if a subject is inebriated.

The convolutional neural network reportedly evaluates thermal infrared images of human faces and determines with more than 90% accuracy if a subject is drunk.

The noninvasive approach to detecting drunkenness could potentially avoid the pitfalls of other approaches that have focused on eye state, head position or functional state indicators — which might be confused by other factors.

The researchers suggest that the convolutional neural network analysis of thermal imaging is less ambiguous and may allow authorities to screen people for inebriation.

The study, Drunkenness detection using a CNN with adding Gaussian noise and blur in the thermal infrared images, appears in the journal International Journal of Intelligent Information and Database Systems.

To contact the author of this article, email