untitled design

Percept system transmits data from cameras 40 times faster than streaming video through AI object selection

OStream, a self-described “object data solution provider for deploying effective artificial intelligence in the physical world,” unveiled the Percept software and data communication system. The development is designed to reduce the cost and time to market of multimedia AI solutions through the use of supporting hardware and software services. Low-cost wireless cameras and sensors are already widely available, OStream says, but the data streams they generate place a heavy burden on customer networks and linearly increase cloud computing costs, limiting practical deployment. Percept solves this problem by integrating “any camera with any AI service.”

Percept system transmits data from cameras 40 times faster than streaming video through AI object selection
Percept system transmits data from cameras 40 times faster than streaming video through AI object selection

OStream Percept consists of three components:

The first is the “Virtual Object Camera”, the firmware of wireless object cameras with a transmission range of 1 km that connects to conventional IoT cameras and forms a software entry point for legacy IP cameras powered by the mains.

The second is the AI ​​Hub server software with AI Objects Router and Data Lake subcomponents to integrate open AI services and provide data storage and search.

The third one is AI Services Market. It aggregates access to over 300 providers of AI models and AI processing services, as well as third-party platforms.

Percept system transmits data from cameras 40 times faster than streaming video through AI object selection

The push for AI media processing at the edge increases hardware and maintenance costs, OStream says, limiting sensor deployment to wall-powered locations with access to existing Wi-Fi. The low-power OStream object codec converts media streams into compact objects, allowing data to be collected and culled for AI at low cost on battery-powered peripherals and then sent over a low power wireless network over long distances.

Wireless connectivity allows Percept-enabled devices to be deployed virtually anywhere and bypass client Wi-Fi, completely avoiding any server-class computing at the edge. In the cloud or locally, OStream deploys its AI Hub to collect candidate data from IoT edge devices, send it to the solution developer’s chosen AI cloud service, and store and search the data in real time. The OStream object codec, low power wireless networking and AI Hub enable computer vision virtually anywhere without relying on the customer’s IT infrastructure.

The OStream Object Codec, which implements Tiny Edge Computing in cameras and other multimedia peripherals, creates lightweight data candidates for AI, thereby reducing the load on video and image transmission channels. The manufacturer estimates that the reduction is as high as 97.5%, so even low-power wireless networks can easily handle the reduced traffic. Ripl Networks MLMesh is used for secure wireless transmission. Endpoints have a range of up to 1 km, which is not available for Wi-Fi or Bluetooth.

The first “Percept Inside” cameras and related devices will become available on March 21st.

Source: ixbt

You may also like

Get the latest

Stay Informed: Get the Latest Updates and Insights

 

Most popular