(Credit: Imagination Technologies Blog) – If you follow this blog closely (and I hope you do), you may have read several articles recently about the great work Imagination is doing to help silicon vendors get to market quickly. One example is the creation of IP platforms that can easily be integrated on chip. In addition, these IP platforms support customizable third party hardware, making them a very flexible solution that addresses many markets.
Earlier last week, Imagination issues a press release announcing that Russian-based ELVEES companies group licensed our PowerVR vision platform, 32- and 64-bit MIPS CPUs and Ensigma RPUs to build next-generation SoCs for a range of video-focused applications.
ELVEES licensed our PowerVR vision IP platform for a new vision processor
Here is what the original announcement said:
The strategic collaboration between ELVEES and Imagination is based on a deep understanding and belief that with the emergence of widespread video analytics, it is becoming impossible for companies to do everything themselves, and that application and domain knowledge are essential. The partnership between ELVEES and Imagination will enable the companies to drive innovation in the video analytics market with more rapid deployment of key technologies and
The ELVEES group includes a fabless design center (ELVEES R&D) and a security and video analytics technologies provider (ELVEES-NeoTek). The group is based on the outskirts of Moscow and focuses on designing software and hardware for computer vision applications, radar surveillance, biometric identification or video analysis systems.
ELVEES headquarters outside of Moscow, Russia
In this article, I’d like to focus on one use case from the list above based on information from Doug Watt, marketing manager for GPU compute at Imagination.
PowerVR imaging framework used for video analytics in retail
Video analytics is a relatively new market and fits under the umbrella of IoT, since it relies on sensor-based processing (i.e. the data coming from the image sensor) and a degree of connectivity (i.e. sending processed analytics information to the cloud for storage/further analysis).
Many video analytics devices so far have re-purposed mobile and embedded chips that were not designed from the ground-up for this specific purpose. Even though there are still incredible savings in performance and power efficiency when running video analytics algorithms on the GPU rather than the multicore CPU, the architecture of mobile chip has different requirements when it comes to video processing and therefore might not be best suited for this market.
An example of a video analytics system built by ELVEES
This brings me back to ELVEES and the PowerVR vision IP platform above. The Russian company is combining the full suite of silicon IP from Imagination (MIPS, PowerVR, Ensigma) with custom, in-house IP designed specifically for vision processing. The programmable ELVEES vision IP cores come from the innovative ELVEES Semantic Video Processing Platform and will be integrated in a family of optimized solutions that will scale from augmented reality on mobile devices to leading-edge solutions for 4K cloud servers.
In addition, ELVEES is using 28nm process technology to build this family of SoCs, a leading node that has been cost-optimized to reduce chip area after place and route.
I am very excited to see more vendors adopting our IP platforms and will continue to present other success stories in the future; stay tuned to the blog!
Interested in the latest news and updates from Imagination? Make sure to follow us on Twitter (@ImaginationPR, @ImaginationTech, @MIPSguru, @MIPSdev), LinkedIn, Facebook and Google+
– Alexandru Voica –