Network Optix, a leading VMS provider, is continuing to expand its offering in cloud-managed services and open AI integration, the company highlighted in an online briefing focusing on APAC partners in Taipei this week.
At the center of the expanded offering is the sixth generation of NX Witness Enterprise, which offers subscription-based upgrades to the company’s perpetually licensed VMS platform NX Witness PRO. This includes expanded scalability—end-users can connect multiple VMS servers via the cloud—as well as the new AI Manager with vast orchestration features.
“Network Optix is continuing to add new default AI features,” said Andy Wen, SEA Regional Director, Network Optix. “But the most important way in which we empower users is that they can upload their own AI models through NX Witness Enterprise’s AI Manager. This is especially interesting for in-house AI developers at Enterprise clients, and those who want to integrate highly specific AI features.”
Wen highlighted the cumbersome way AI developers usually get their own models to run on a typical VMS: Once the model is ready, they need to write software around it and integrate that software with the VMS. AI training might take as little as two weeks, but the whole process might take up to six months, Wen stated.
AI Manager enables developers to skip several steps. “Once the model is ready, you upload it to the VMS and you can use it immediately,” Wen explained.
AI Manager also comes with a proprietary AI model library and is bundled with Enterprise. It was designed to run on on-prem devices with a slim footprint in mind. It is compatible with all AI models in ONNX format, which is common for open-source models from GitHub and similar platforms. Licensing is per camera channel, not per model, allowing users flexible mix-and-match approaches.
Meanwhile, the Model Chain feature of AI Manager helps users orchestrate multiple models. Wen gave the example of object detection: Users can run one model to detect, for example, the shape of an object, and send positives to another model that counts them. While many off-the-shelf AI models can classify objects along a large number of vectors, the Network Optix approach enables integrators to build highly specific systems.
“Take for example a sushi restaurant,” Wen said. “We have a case where a customer built an AI-based sushi recognition system that alerts the kitchen when a certain product—let’s say tuna hand rolls—is running low on the conveyor belt.”
Inspired by the large number of AI use cases, Wen said Network Optix is planning to open an “AI Shop” for integrators, resellers and end users, where they can upload and license their models to others.
Scalability
Aside from AI options, Enterprise enables greater scalability. While Network Optix continues to focus on on-prem architectures (with the possibility that users move the VMS to a third-party cloud), Enterprise helps them build a bridge between multiple servers—at the same site or at multiple sites, for example across multiple branches of a chain store.
Wen said that Network Optix recommends 10 as the maximum number of servers, as unstable Wi-Fi or 4G connections might become a bottleneck in larger architectures. However, he added that up to 100 servers had been connected and run without issues during lab tests at Network Optix so real-world architectures with far more than 10 servers are doubtlessly possible if the operator can provide suitable conditions.
While Network Optix and SIs of its partner network license Enterprise through a subscription-based model, the company is not moving away from its “perpetual-license” approach, Wen said, adding that rumors that Witness will become subscription-based are false.
“If a user licenses Enterprise, or a certain subscription feature within Enterprise, for let’s say two years, and they don’t renew their license in the end, they simply fall back on NX Witness PRO,” Wen said. “Their system remains functional without Enterprise features.”
One VMS, many options
Wen highlighted that NX Witness offers a great number of customization features, as well as built-in options that help address real-world pain points.
One major focus is greater situational awareness. Wen said that for example fisheye cameras have been trending in the past few years as they give users 360-degree overview, with the downside that objects appear deformed or turned on their heads in the primary stream. NX Witness therefore includes Fisheye Dewarping, which enables users to click on an object in the fisheye live-stream, and the VMS zooms in and “dewarps” the object to make it appear natural. With Network Optix, users can also add fisheye lenses to legacy cameras, giving them full situational awareness from every vantage point—as well as the VMS-based dewarping feature.
Wen also highlighted that NX Witness allows users to create virtual maps of their premises. If they, for example, observe a person in a live stream, they can click “next camera” and the VMS jumps to the next position along the person’s path.
Enterprise also unlocks AI Maps, which Wen said is especially interesting for operators of vast smart city or traffic control systems. Running on the feeds of vast camera networks—in Wen’s demo it was Taiwan-wide traffic monitoring—AI Maps plots AI-recognized objects as live, moving icons—people and different types of vehicles for example—using metadata from AI Manager or in-camera AI.
With AI Maps, operators can address issues such as traffic congestion much more efficiently, Wen explained.