Renesas, along with Fixstars, is taking deep learning designs built around its R-Car SoCs to the cloud to facilitate instant initial evaluations when selecting chips for ADAS and autonomous vehicle designs.
One more chipmaker is taking its semiconductor design solutions to the cloud in a bid to reinvent the developer experience while easing evaluation tasks for deep learning-based automotive designs. In collaboration with software house Fixstars, Renesas is establishing the Automotive SW Platform Lab next month to help optimize software development for advanced driver assistance systems (ADAS) and autonomous driving applications. Fixstars specializes in software solutions built around multi-core CPU/GPU/FPGA acceleration technology.
Automotive developers are increasingly turning to deep learning for new ways to enable cameras and sensors in ADAS and autonomous driving systems. However, software development and delivery are a major pain point for automotive system developers, especially for deep-learning designs that have been predominantly built for consumer and server applications.
Figure 1 Automotive SW Platform Lab will provide evaluation services for deep-learning camera and sensor designs built around the R-Car chips. Source: Renesas
Renesas, along with Fixstars, is taking deep learning designs built around its R-Car system-on-chips (SoCs) to the cloud to facilitate instant initial evaluations when selecting chips for ADAS and autonomous vehicle designs. Renesas will employ Fixstars’ cloud-based device evaluation environment, GENESIS, for the early development of deep learning-based automotive applications.
From SDK to cloud
It’s worth mentioning that Renesas has already made available the R-Car Software Development Kit (SDK) to enable quicker and easier software development and validation for automotive designs. The software platform, which comes in a single package, is built for rule-based automotive computer vision and artificial intelligence (AI) functions. The SDK includes a full set of software samples, popular CNN networks, and application notes.
Figure 2 The e² studio in the SDK has added new features, including support for bus monitoring and debug functionalities for image processing and deep-learning subsystems. Source: Renesas
While design engineers commonly use evaluation boards and accompanying software to evaluate devices, technical expertise is still required to build an evaluation environment. On the other hand, a cloud-based evaluation environment, like the one Renesas is offering, doesn’t require specialized technical expertise.
Next, a cloud platform allows developers to confirm the processing execution time in frames per second (fps) and recognition accuracy percentage of CNN accelerators on sample images using generic CNN models like ResNet and MobileNet. Moreover, designers can use the cloud platform to confirm evaluation results in tasks like image classification and object detection with the option to use their own image or video data. That, in turn, enables engineers to determine if the chip is suitable for system design.
Satoshi Miki, CEO of Fixstars, points out that the cloud platform for automotive SoCs is also crucial because, after developing a deep-learning application, it’s not possible to maintain high-recognition accuracy and performance without constantly updating it with the latest learning data. The cloud-based design platform allows developers to continuously update learned network models to maintain and enhance recognition accuracy and performance.
When it unveiled the SDK for R-Car chips in September 2021, Renesas vowed to follow up with a virtual platform. Here it comes with a cloud-based evaluation toolset, promising to simplify the software development for automotive designs serving ADAS and autonomous vehicles in passenger, commercial, and off-road environments.
This article was originally published on EDN.
Majeed Ahmad, editor-in-chief of EDN and Planet Analog, has covered the electronics design industry for more than two decades.