Detection of Armed Assailants in Hostage Situations- A Machine Learning based approach

Innocent persons may be captivated by a criminal abductors and may threaten the person's employer, government, relatives, etc. To exploit unlawfully for their advantage after expiration of an ultimatum. Abducting people and threatening is a criminal act or act of terrorism. It is important to distinguish the assailants from the hostages. The rescuer has milliseconds to distinguish and any slight mistake in identifying a person might cost the lives of all the innocent hostages in that situation. To help the rescuers with this a model is developed in this research work that can identify if a person is an assailant or the hostage in a hostage situation. The proposed model takes the real time video as input and detects the assailant by making a rectangular bounding box on his/her face. YOLO (You only look once) and OpenPose algorithms are used in the proposed system and both algorithms are object detection systems targeted for real-time processing for both classification and localizing the object using bounding boxes and Heat maps.