promiseQ - Big FuE  (Brandenburgischer Innovationsgutschein) Project

promiseQ’s AI model for detecting false alarms

By combining the accuracy of AI with a human touch, we are making video surveillance more reliable than ever. promiseQ now has its own AI model that is able to detect people and moving objects, thanks to assistance from the Big FuE (Brandenburgischer Innovationsgutschein) project.

With this new cutting-edge technology, promiseQ can now reduce false alarms on a larger scale and help security clients focus on what's important – protecting people and property from real threats.

Founded in 2021, promiseQ now has a functional false alarm filter - Threat Detect - for IP cameras used in video surveillance systems. promiseQ initially investigated whether markers provided in alarm testing were sufficient to train a false alarm AI. The company has been successful in doing so.

Threat Detect now uses the most advanced neural network algorithms to classify actions within video streams. This false alarm filter ensures close integration of AI and human intelligence on one platform and does so in real-time.

Our false alarm filter works with the help of human-in-the-loop configuration. Human-in-the-loop is a form of AI modeling that combines the best of artificial intelligence with the strengths of human action.

Thanks to human input we’ve been able to create a false alarm filter with the highest possible accuracy for determining emergencies from harmless events.

The process of identifying false alarms and communicating them to human operators has up until recently been distributed across several systems or even many companies.

Threat Detect tests have shown that the promiseQ AI can increase alarm filtering rates by 15% up to and over 95%. 

A typical reduced false alarm rate of the promiseQ AI is often above 95%, depending on the camera, the use case, and the data distribution. This means that out of 100 false alarms, the Threat Detect engine correctly filters out approximately 95.

Individual pilot projects where customers can test out Threat Detect have been successful and resulted in client uptake of the product.