Edge Impulse Partners with Arm to Integrate with the Arm Keil MDK for Easier, Faster Edge AI

Edge Impulse has announced that machine learning (ML) models built on its platform can now be integrated directly into Arm's Keil Microcontroller Development Kit (MDK) — making it easier to take a design from prototype to production.

"Serving embedded engineers with new tools that help them get started with AI [Artificial Intelligence] and ML projects quickly is an important priority to us, and working with Arm to support Keil Studio enables us to impact one of the largest communities in edge development," says Jan Jongboom, Edge Impulse co-founder and chief technology officer, of the partnership. "It's our hope this partnership will lead to greater utilization of AI to improve the critical firmware powering industrial, IoT [Internet of Things], and wearable environments."

"ML and AI functionality are becoming more widely adopted for edge use cases, and embedded developers need tools to navigate these complexities," adds Reinhard Keil, senior director of embedded technology at Arm, of the pair's work together. "Keil MDK is the most trusted and widely deployed embedded development environment in the ecosystem. Integrating Edge Impulse’s edge AI tools within it brings additional capabilities to a vast community of developers enabling them to build sophisticated edge AI firmware faster than ever on Arm."

The integration between Edge Impulse Studio and the Keil MDK comes courtesy of the open Common Microcontroller Software Interface Standard (CMSIS) project Open-CMSIS-Pack, which standardizes packaging and distribution of software components. Edge Impulse projects built with the company's latest software development kit (SDK) can be deployed as an Open-CSMIS pack and imported into any Kiel MDK project, the companies say, with a pack manager ensuring that the SDK remains up-to-date for each release.

Projects integrated into the Keil MDK can also be optimized for Arm platforms using the Edge Impulse Edge Optimized Neural (EON) compiler and EON tuner auto-ML tool, the company has confirmed, promising savings up to 70 per cent in RAM usage and 40 per cent in flash usage on top of efficiency and performance tweaks available in the Arm compiler.

More information on the integration, and a getting-started guide, can be found on the Edge Impulse blog.