Check out all the on-demand sessions from the Intelligent Security Summit here.

Chip designer Ambarella has announced two new chips for automotive cameras and advanced driver assistance systems (ADAS) based on its CVflow architecture for artificial intelligence processing. The Santa Clara, California-based company unveiled the CV22FS and CV2FS automotive camera system-on-chips (SoCs) with CVflow AI processing and ASIL-B compliance to enable safety-critical applications.

Ambarella will also demo applications with its existing chips — as well as a robotics platform and Amazon SageMaker Neo technology for training machine-learning models — at CES 2020, the big tech trade show in Las Vegas this week.

The company, which went public in 2012, started out as a maker of low-power chips for video cameras. But it parlayed that capability into computer vision expertise and launched its CVflow architecture in 2018 to create low-power artificial intelligence chips. Now it has 760 employees and is competing with the likes of Intel and Nvidia, though with a focus on low-power applications.

Senya Pertsel, senior director of marketing at Ambarella, said in an interview with VentureBeat that customers such as German software maker Hella Aglaia chose CVflow because of the low-power consumption.


Intelligent Security Summit On-Demand

Learn the critical role of AI & ML in cybersecurity and industry specific case studies. Watch on-demand sessions today.

Watch Here

“If you look at the market, you see that the form factor is very constrained and yet it has to work in extreme temperatures,” Pertsel said. “And you have to detect things like cyclists, pedestrians, and so on. You have to detect across extended distances … and see things like cross traffic, which requires a wide field of view. So you are driven to higher resolution cameras. You need more and more processing power for complex AI algorithms, with the constraint on power consumption.”

Above: Senya Pertsel is senior director of marketing at Ambarella.

Image Credit: Ambarella

Both new chips target forward-facing monocular and stereovision ADAS cameras, as well as computer vision ECUs for Level 2+ and higher levels of autonomy for self-driving cars.

Featuring extremely low power consumption, the CV22FS and CV2FS make it possible for car makers to meet performance requirements within the power consumption constraints of single-box, windshield-mounted forward ADAS cameras.

Other potential applications for the processors include electronic mirrors with blind-spot detection (BSD), interior driver and cabin-monitoring cameras, and around view monitors (AVM) with parking assist.

The two new SoCs are the latest additions to Ambarella’s successful CVflow SoC family, which offers automotive companies and software development partners an open platform for differentiated, high-performance automotive systems.

ZF, a global automotive supplier of systems for passenger cars and commercial vehicles, is working with Ambarella on viewing and sensing systems. Aaron Jefferson, vice president of ADAS product planning at ZF, said in a statement that the companies are working on surround view, e-mirror, and computer vision technologies.

Hella Aglaia, a maker of intelligent visual perception software, has worked with Ambarella’s CVflow processors over the past year. The customer chose CVflow SoCs due to their ability to deliver extremely high computer vision processing performance with very low power consumption, said Kay Talmi, managing director at Hella Aglaia, in a statement.

The CVflow architecture in the CV22FS and CV2FS provides computer vision processing in 8-megapixel or higher resolutions at 30 frames per second for object recognition over long distances and with high accuracy. The SoCs each include a dense optical flow accelerator for simultaneous localization and mapping (SLAM), as well as distance and depth estimation. Ambarella says its multi-channel high-speed sensor input and image signal processing (ISP) pipeline provides the necessary camera input support, even in challenging lighting conditions. CV2FS also enables advanced stereovision applications by adding a dense disparity engine.

Ambarella will show its CVflow SoC family to select customers and partners during CES 2020. Demonstrations will include Hella Aglaia’s deep learning ADAS algorithms and Ambarella’s EVA (Embedded Vehicle Autonomy) self-driving prototype vehicle. Ambarella will also demonstrate a range of applications from other key partners running on the CVflow engine.

CV22FS and CV2FS are scheduled to sample to Ambarella customers in the first half of 2020.

CES 2020 demos

Above: Ambarella’s CV2FS chip handles AI processing in cars.

Image Credit: Ambarella

Among the CES 2020 demos, Mercedes-Benz will demonstrate its CV2-based Cargo Recognition and Organization System (CoROS). A camera assistant in the cargo space automatically recognizes registered parcels using barcodes and the symbols on the outside of the parcels. This process is done in fractions of a second, replacing manual, time-consuming scanning and sorting of each shipment.

And Hella Aglaia will feature its latest suite of deep learning ADAS algorithms, including multi-class object detection, detection of driving area limitations, depth estimation, and classification of traffic lights and traffic signs. Running on a single Ambarella CV22 CVflow SoC, this ADAS platform supports the development of single-box, forward-facing ADAS cameras.

South Korea-based StradVision will demonstrate its suite of front ADAS and driver monitoring system (DMS) algorithms running on a single CV22. Connected to an 8-megapixel front-facing camera and an additional interior-facing camera, this system will be installed and running in a vehicle.

Israel-based EyeSight’s driver monitoring solution (DMS) will be shown on a system with three cameras. In this demonstration, Ambarella’s CV25 will simultaneously process a monochrome driver-facing camera and two RGBIr in-cabin cameras (each with a different field of view).

And Israel-based Brodmann 17’s ADAS solutions suite will showcase the company’s deep learning algorithms, which include vehicle detection, distance estimation, and real-time forward-collision warning, also running on a CV22 SoC.

VentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.