Feature Articles
Xjera Labs uses NVIDIA's Tesla P4 in its surveillance solutions
By Ian Chee - 2 May 2017,5:16pm

Xjera Labs uses NVIDIA's Tesla P4 in its surveillance solutions

By now, you’ll probably already know that NVIDIA is making major progress in the world of artificial intelligence (A.I.) and Deep Learning. The company has the hardware to make it happen, such as the Drive PX 2 and the Tesla P100, and the company also provides training to those who would want to use its hardware to run their own A.I. development and training. We’ve spoken to a representative of a company that uses NVIDIA tech in its A.I. applications to learn more on the many benefits NVIDIA hardware has brought to the company.

That person was Grace Wei, VP, Project Management at Xjera Labs, a Singapore-based company that deals in Intelligent Video Analytics, a less intimidating term to describe surveillance. As a company that delivers surveillance services to its clients, Wei starts off immediately by mentioning the production quality of NVIDIA’s solutions, namely the Tesla P4. For other companies that require surveillance, it would mean that the hardware involved would have to work continuously for extended periods of time. Due to the nature of GPUs and their inherent difference with CPU architecture, it makes using GPUs a lot more cost effective. And to meet all these needs, Xjera Labs found that the Tesla P4 is the best solution that could deliver on all these aspects.

NVIDIA Tesla P4.

The ease of scalability and reconfigurability of the Tesla P4, with the added benefit of NVIDIA’s GPU support of a multitude of Deep Learning frameworks, gives Xjera Labs the flexibility that it needs to not only differentiate itself from the competition, but also gives them the flexibility and possibility to do more with their Intelligent Video Analytic A.I.

Wei also said that Xjera Labs’ cameras need very little fine-tuning, if any at all, thanks to its own Deep Learning algorithm developed with the help of NVIDIA engineers. And with customer feedback, the already accurate surveillance system can be made even better, depending on the situation, and over time Xjera Labs is able to offer more customized solutions to its clients.

Xjera Labs' Intelligent Video Analysis solution making out individual persons in a crowd. <br>Image source: Xjera Labs.

In the real world, Xjera Labs’ video analytics solution does pretty much as described by Dr. Simon See, Chief Solutions Architect, NVIDIA. The algorithms allow for the surveillance solution to recognize people and vehicles. On the former front, capabilities include detection and description, segmentation and facial recognition. But more than that, Xjera Labs’ surveillance solution also recognizes behavior and situations where people are in motion, which allows it to send alerts whenever it detects, say, a brawl, or when someone is smoking in a non-smoking area.

Here's an example of Xjera Labs' A.I. detecting a brawl. <br>Image source: Xjera Labs.

For vehicles, the surveillance system can detect the make and model of a car, its color, its license plate number, and if it is illegally parked in its current position. Being able to detect motion, it is also possible for the A.I. to determine if someone is reverse-driving, making illegal u-turns, or any other dangerous behavior behind the wheel.

Detection of behavior in front of a shelf. <br>Image source: Xjera Labs.

For inanimate objects, Xjera Labs intends to train its A.I. solutions to recognize other inanimate objects. This allows the company to expand and enter other markets, and one of the examples that was given to us was smart retail, where an A.I. can recognize and keep track of items that are taken off a shelf in a store and help with keeping track of stock inventory – much like Amazon Go.