UCF stands for University of Central Florida. This is a large dataset containing 128 hours of video. It consists of 1,900 continuous, unedited surveillance videos, covering 13 types of abnormal behaviors in real life, namely abuse, arrest, arson, assault, road accident, burglary, explosion, fighting, robbery, shooting, theft, shoplifting, and vandalism. These abnormal behavior videos were selected because they have an impact on public safety.

This dataset has two uses: one is general abnormal behavior detection, which is to compare one set of abnormal behaviors with another set of normal activities to draw conclusions; the other is to identify the types of abnormal behaviors.