Action recognition examples
MIDS Capstone Project Fall 2019

Bittah-Ninja Violence Detection

Security professionals spend hours with their eyes glued to security camera feeds because they are judged on response times to incidents happening "on their watch" (and also because responders must have "who, what, where & when" details in order to respond quickly and appropriately). Envisioning an AI system that could send de-escalation resources immediately upon recognizing violence, the Bittah-Ninjas finely-tuned a Temporal Segmentation Network to flag video footage of people throwing punches. Within venues where weapons-screening happens pre-entry (such as airports, ships, prisons, and some stadiums, schools, and conference centers), automated punch detection could build confidence in our approach and capabilities. Further investment and development efforts would be directed toward detecting those concealing or drawing lethal weapons in public areas. Enabled with real-time filtering for those holding valid, RFID-embedded concealed-carry permits, such a system could help prevent - or at least minimize the damage of - future mass shootings.

Man about to throw a punch in a stadium
Man about to throw a punch in a stadium

Last updated:

December 11, 2019