Integration of a Machine-Learning based Threat Identification with VICE

(VICE) to collect faces and identify them in large photo and video collections.   In the proposed work we will extend this capability to incorporate vehicles, persons, and weapons recognition into the VICE so that people, weapons, and vehicles can be correlated and identify as friend of foe (i.e. non-threat and threat).   Cybernet built for the Army Research Office (and later the Air Force and the Joint IED Organization), a learning-based computer vision system that incorporates (1) hand or manual and machine learning capabilities, (2) truth data collection and processing, (c) location and history retention, and (d) animals and threat recognition learning and automatic tagging.  This system has been applied to generic recognition and tracking of vehicles, person & carried items, generically potential threats, and species of animal for environment reporting and identification counts. Our system includes tools and Machine Learning (ML) methodology (and existing vehicle, person, and carried item recognition codes) that should readily port to the VICE environment for an immediate proof of concept result in Phase I and can be extended through processing and assessment of larger Phase II training sets to continuously improve recognition rates and threat-Identify pairing performance.