Semiconductor supplier NVIDIA in March introduced a new version of its Jetson embedded computing platform that enables edge devices like surveillance cameras to perform AI object recognition.
NVIDIA’S Jetson TX2 doubles the performance of the previous platform in the product line, the TX1, according to TechCrunch. Specifically, the TX2 includes an Nvidia Pascal-based graphics processing unit (GPU) chip that the company said provides an exponential leap in capabilities compared to the previous-generation Maxwell GPU used in the TX1.
Pascal adds new algorithms specifically designed for AI deep-learning applications, according to NVIDIA. These algorithms allow devices to conduct real-time deep-learning inference, a technique that allows AI systems to comprehend new inputs based on previous training.
The capabilities of the TX2 are suited for use in smart surveillance cameras that are equipped with AI processing hardware and software. Such cameras are designed to detect objects and events that could represent a security risk.
For organizations that engage in video surveillance, AI processing conducted at the edge is a highly desirable feature. AI object recognition technology can dramatically enhance surveillance cameras, improving their effectiveness while reducing the amount of manpower required to monitor the video output.
However, many current AI solutions require cameras to send their video data out to the public cloud for processing. This raises various concerns, including issues related to the speed of reaction to events detected by the camera.
To conduct effective video surveillance, organizations need to react quickly to occurrences such as safety or security threats. Public cloud-based AI technology is simply too slow for this task.
The arrival of the Jetson TX2 and similar platforms promises to proliferate the use of speedy AI object recognition technology in surveillance cameras and other edge devices.
Tyler Schulze is vice president, strategy & development at Veritone. He serves as general manager for developer partnerships, cognitive engine ecosystem, and media ingestion for the Veritone Platform. Learn more about our platform and join the Veritone developer ecosystem today.