South Korea’s Electronics and Telecommunications Research Institute (ETRI) has introduced Dejaview, an AI system designed to detect and predict crimes in real-time. It analyzes CCTV footage and utilizes machine learning to examine patterns and identify signs of potential crimes.
The AI system includes a detection feature capable of identifying high-risk individuals and analyzing their actions and body movements. It is designed to alert authorities when suspicious activities are detected by comparing them with past criminal data. This strengthens the authorities’ proactive crime prevention capabilities. It also features a Predictive Crime Map (PCM) to identify high-risk zones, enabling it to predict potential incidents such as violent acts, traffic accidents, and fires. Authorities can then actively monitor those high-risk zones more efficiently through real-time CCTV feeds.
The system was trained on a vast dataset of over 32,000 CCTV clips documenting various incidents across a three-year span. The AI has assimilated the ability to discern patterns from this dataset and now applies this knowledge in real-time situations. Dejaview has demonstrated an accuracy rate of 82.8%.
