Lumana AI Inc., maker of a cloud-based security platform with image recognition capabilities, is emerging today with $24 million in seed funding.

The company says its technology transforms any security camera into an AI-powered device capable of identifying complex events and risks. Its platform is built on a distributed hybrid cloud architecture that integrates cloud and on-premises resources and dynamically chooses processing options for optimal performance. Embedded deep learning artificial intelligence models continuously improve over time, and customers can define what types of events should generate alerts.

“The customer must install Lumana Core, which includes an AI engine and video processor,” said founder and Chief Executive Sagi Ben Moshe, who was previously chief incubation officer at Intel Corp. after it acquired the 3D imaging company Invision Ltd. he founded. “The Core processes the video footage and runs AI on the edge for greater reliability and speed. It also interacts with the cloud for communication with other Lumana devices and with our video management application.”

Lumana cited Forbes Business Research, which projects that the video analytics market will more than quadruple from over $8 billion this year to almost $38 billion by 2030.

Accessible anywhere

The company’s video security platform includes the AI engine and video processor, video management software accessible through a web browser or mobile application and security cameras, although any internet-connected camera will work.

Administrators can manage an unlimited number of cameras and locations from a single console, which offers advanced user management and permissions, sharing capabilities, video playback, system health dashboards, alert monitoring and other features. Alerts are typically generated in less than a second.

Lumina enables natural language customized alerts to be triggered in real time, and its self-learning features reduce false alarms over time. The model can be configured for specialized alerts such as car break-ins, threatening behavior or a worker’s failure to wear personal protective equipment on a manufacturing line.

It can recognize events and elements such as faces, clothing types, vehicle attributes and license plates, proximity, vehicle speed, doorway “tailgating,” loitering, tampering, unidentified people and activity heatmaps.

Fast search

One of the platform’s defining features is its ability to search millions of hours of video across locations with results returned in seconds. Users can specify multiple search parameters, such as vehicle attributes, to locate footage.

“We are running AI on the edge and, if needed, continue to run it on the cloud,” Ben Moshe said. “We then save the AI’s output in data structures that let us handle big data queries much faster than existing systems. This approach lets customers quickly search video footage for complex scenarios.”

Ben Moshe said parameters can be combined to pinpoint exact footage. “For example, two people are together, one wearing a black hoodie and blue jeans and another wearing a white shirt with red shoes. These people were last seen in a white Tesla with a license plate that includes XYZ,” he said. “Our system will return all scenes that match those criteria within seconds.”

The technology also powers what Lumana calls “natural language alerts,” such as “alert me when an unknown person is detected at Building A, 30 minutes later at Building B, and then the next day at Building B again.”

Customizable dashboards enable users to monitor specific activities, such as unauthorized access occurrences, alert types by building or location, safety violations, near-miss accidents and space utilization.

Norwest Venture Partners LP and S Capital VC Management Ltd led funding.

Image: Lumana