Lumana's focus on vertical applications for AI video surveillance platform

Lumana, the enterprise video security system that turns standard IP cameras into AI-powered monitoring agents, is extending its platform beyond physical security into retail, healthcare, logistics and smart city applications, according to the company's Principal Product Manager, Orit Dolev.

The firm's proprietary technology, VIA-1, is designed to interpret footage from cameras that organisations already have installed, rather than installing new hardware. Call it camera-agnostic.

Dolev said the model is tuned to each individual camera it monitors, accounting for its viewing angle, lighting, and the objects and people typically captured within its frame. During initial deployment, each camera establishes a baseline of normal activities in the surrounding area it is monitoring so that during live operations, cameras can understand their environments and highlight exceptions to routine behaviours.

"The way our model works is it's specialised and refined per camera. It's not just a general model that is the same for every one of the platform's cameras. For every camera, it has its own unique angle, [taking into account] the lighting in that area, the elements in the scene, objects and people, and our model has the capability to adapt and refine based on that specific camera and scene that it already has so much data on," said Dolev.

The platform lets operators search footage using free-text queries, such as a description of clothing, and returns matching results across every connected camera to the platform's central dashboard.

She added that the emphasis on simple, intuitive search was a deliberate design choice, based on customer feedback that competing systems could be complex and difficult for operators to use on a daily basis.

Objects or individuals identified in a search can then be tracked across their movement through a site. The system is also designed to interpret behaviour within a scene, including recognising when a person appears to be carrying a weapon, Dolev said.

During camera-customised training, Lumana incorporates data sampling directly into the curation pipeline. Once deployed, the company perpetually audits production models against updated datasets and real-world scenarios to catch drift before it impacts performance.

Retail innovations

In retail, Lumana positions its technology as a tool that can address additional industry tasks such as stock availability and store analytics rather than solely theft prevention.

"In supermarkets, the average shelf availability sits around 90 per cent, meaning one in 10 times a shopper tries to buy an item like milk or bread, it's not on the shelf. That's a growth problem," Dolev said.

With the right camera angle, the platform continuously monitors shelf conditions and alerts store managers when stock runs low, allowing staff to restock before a sale is lost to a competitor, according to Dolev.

The same cameras used for stock monitoring can also be applied to loss prevention, flagging suspicious point-of-sale transactions and potential theft in real time.

Smart cities and campuses

Government bodies and city authorities represent one of Lumana's largest customer segments, alongside physical security more broadly, according to Dolev. Applications include identifying security threats, suspicious activity, traffic-related incidents and infrastructure issues. 

Unlike smaller retail deployments, where a manager may only receive an alert when the system flags an event, city-level deployments generally involve a dedicated monitoring team, Dolev said. That team reviews only footage the system has identified as significant, based on alert criteria agreed with the customer, rather than monitoring general video feeds. 

Lumana said it could not name any governments or smart cities using the technology.

Educational institutions are embracing the autonomous systems as well, with the University of Miami and Geneva College using the technology within their operations.

"Alerts are defined together with a customer to better understand what are their needs and what alerts they want to receive. For some alerts, it's only popping up the video clip. For some of them, it's calling for emergency care. So there is a different kind of dispatching that we can do with the system, including integrating with access control to lock doors if something happens that requires that kind of action. The platform is very versatile and very much tailored to the customer's needs," she said.

Other new markets

Physical security remains Lumana's largest customer segment, spanning small and medium-sized businesses through to large enterprises. Dolev said existing customers have increasingly sought additional applications for the technology beyond security and safety, prompting the company's expansion into new verticals such as quick-service restaurants.

The company is also expanding into manufacturing, logistics and warehousing surveillance, where the company says its cameras can automate detection of personal protective equipment and flag workplace safety breaches.

In healthcare, applications include detecting falls, identifying aggressive incidents that require intervention, and monitoring restricted or sensitive areas of hospitals.

"We're turning passive video into an active intelligence layer, analysing over a billion images daily across more than 50,000 cameras, to prove the future of physical enterprise operations is already here."