AI detects weapons in video footage
The American company Athena Security uses AI techniques to analyse a live video stream and recognise an assailant carrying a firearm, or acting aggressively...
The American company Athena Security uses AI (artificial intelligence) techniques to analyse a live video stream and recognise an assailant carrying a firearm, or acting aggressively. The system can be applied to an existing video surveillance installation where Athena reportedly will link to the video feed via the camera IP address and use its Cloud-based resources. The company claims they can also install a full system incluAIfiding all the hardware as required. When a hazardous situation is identified the system issues vocal warnings which can help defuse a potentially lethal conflict event. It can also be configured to bar access to areas when a threat is identified at a particular location and will automatically alert authorities and police via an app.
The system is already installed in a number of schools and college campuses in the US, a country that has witnessed some tragic incidents recently. It uses powerful Nvidia graphics cards together with AI computer vision to analyze the 30 fps video stream and detect the presence of dangerous objects and threatening body gestures. Sophisticated AI recognition systems such as this are prone to ‘false positive’ detections but the system is able to feed back errors to improve accuracy. The company claims that the system demonstrates a better than 90% success rate at identifying weapons appearing in a live video stream.
The system is already installed in a number of schools and college campuses in the US, a country that has witnessed some tragic incidents recently. It uses powerful Nvidia graphics cards together with AI computer vision to analyze the 30 fps video stream and detect the presence of dangerous objects and threatening body gestures. Sophisticated AI recognition systems such as this are prone to ‘false positive’ detections but the system is able to feed back errors to improve accuracy. The company claims that the system demonstrates a better than 90% success rate at identifying weapons appearing in a live video stream.
