Credit: University of Granada

Researchers working at the University of Granada (UGR) have developed artificial intelligence programming that automatically detects in real time when a gun is drawn in a video.  It speaks to the times we live in that technology like this would sound so practical and valuable.   Siham Tabik, Roberto Olmos and Francisco Herrera Triguero scientists in the Department of Computational and Artificial Intelligence Sciences at the UGR, have taken the AI system they have designed into a prototype computer system that analyzed low-resolution videos from YouTube and action movies from the ’90s.

Prototype Details

The AI based computer system analyzed at a rate of five frames per second in real time and showed to be capable of detecting guns with high precision, at just over 96.5%.  The AI algorithm which uses deep learning models triggered a warning light when a gun was identified.  Deep Learning is a major breakthrough in the identification and classification of objects. It uses algorithms modeled like the structure and function of the brain called artificial neural networks to create code like taxonomies.

Benefits of This System

This AI based computer system is fast and can run on a computer with moderately high capacities.  It can be deployed without direct human supervision inexpensively and it can be easily integrated in with an alarm system and a video camera security system.  The gun detection system has many potential practical applications, like monitoring violent content in which guns appear in videos uploaded on social networks.   And adding an extra layer of security in building, malls, airports or anywhere large crowds gather.  Often in these crowd scenarios if there are limited points of entry, there may be check points and even metal detectors that can be proactive in that they can find hidden weapons and not wait till someone pulls a gun.  As the authors of this work point out these systems are not practical in many security scenarios where they are too expensive, cause false alarms and bottlenecks waiting to get through.