Recogni, the startup focused on AI-based vision perception processing for autonomous driving platforms, will introduce at CES 2023 what it says is the industry’s highest-performing compute platform for autonomous mobility. Scorpio is a 1000 TOPS (tera operations per second) inference solution, competing with offerings from Qualcomm and Nvidia—the latter launching its 1000-TOPS Drive Atlan system-on-a-chip at its GTC event in April 2021.
Rather than beefing up the central compute power of an SoC, Recogni wants carmakers to place its 1000-TOPS processor at the vehicle’s edge, right next to its CMOS (complementary metal oxide semiconductor) image sensor. The vision-inference chip enables “superhuman” object detection accuracy. The company touts up to 300 m (984 ft) in real-time detection under various road and environment conditions, and the ability to process multiple streams of ultra-high resolution and very-high-frame-rate cameras.
“Recogni’s purpose-built architecture is a unique approach to AI perception, allowing customers to perceive the environment in high resolution with very low latency and low power,” said Marc Bolitho, Chief Executive Officer at Recogni. “This is a game-changer for OEMs and suppliers looking to add new, powerful ADAS and self-driving features to new vehicles. Recogni’s unique approach in perception processing is truly the first of its kind and enables customers to deploy safe autonomous driving functions as well as extending the range for electric vehicles.”
Founded in 2017, Recogni has offices in San Jose, CA, and Munich, Germany. Lead investors are GreatPoint Ventures, Celesta Capital, Mayfield, and DNS Capital—as well as notable automotive OEM and Tier 1 suppliers including BMW iVentures, Toyota Ventures, Bosch, Continental, Forvia, and FluxUnit Ams Osram Ventures.
Currently being evaluated by several “top tier” automotive manufacturers and suppliers, Recogni says its solution is not only 10-20 times more power efficient than competing solutions but also it enables the flexible design of autonomous driving vehicle stacks and minimizes the impact on driving range. In addition, with such a short processing time of less than 10 ms, the ECU (electronic control unit) has more-than-ample time for making driving decisions. High compute capacity, efficient processing, low latency, and low power consumption are key aspects of Recogni’s platform.
“Vision is fundamental to accurate perception processing and essential to autonomous driving platforms,” said RK Anand, Founder and Chief Product Officer at Recogni. “From the beginning, we took a unique approach of processing high-resolution images at the edge to achieve near-perfect object detection and classification and enable autonomous driving stacks to make better driving decisions. Scorpio can process multiple 8-MP streams at 30 fps in less than 10 ms using only 25 W. That’s a performance order of magnitude greater than anything else on the market and, we believe, will help to accelerate autonomous driving to become a reality.”
At CES, the company will demonstrate why “full-frame” vs. “region-of-interest” capture is better, the benefits of stereo 8-MP (vs. 2-MP) data, and low latency while providing 1000 TOPS at 10 W.