“ST Unveils World’s First MCU AI Developer Cloud to Work Seamlessly with STM32Cube.AI Ecosystem”

“ST’s goal is to help developers and data scientists solve their pressing challenges while developing edge AI applications faster and more conveniently,” said Remi El Ouazzane, President, Microcontroller and Digital IC Group, STMicroelectronics. “This new tool will save effort and cost by enabling remote benchmarking of STM32 hardware models through the cloud.”

STMicroelectronics (NYSE: STM), a global semiconductor leader serving customers across the spectrum of electronics applications, today announced the industry’s first set of tools and services designed to support hardware and software decision-making and to quickly and easily bring edge AI technologies to market.

The STM32Cube.AI developer cloud provides access to a comprehensive set of online development tools built on the STM32 microcontroller (MCU) family. With this, ST emphasized that it is continuously expanding solutions to support embedded AI developers and data scientists.

To meet the growing demand for edge AI-based systems, the STM32Cube.AI desktop front-end includes resources for developers to verify and generate STM32 AI libraries optimized for trained neural networks. The online version of the tool, the STM32Cube.AI developer cloud, is further enhanced with a range of industry-first features.

The online interface generates optimized C code for STM32 microcontrollers without the need for pre-installation of software, allowing data scientists and developers to develop edge AI projects by taking advantage of the proven neural network optimization performance of STM32Cube.AI.

ST says access to the STM32 Model Zoo and a repository of trainable deep-learning models and demos can accelerate application development, and example use cases available at launch include human gesture detection for activity recognition and tracking, computer vision for image classification or object detection, and audio event detection for audio classification. The examples are hosted on GitHub and support automatic generation of STM32-optimized “getting started” packages.

The world's first online benchmarking service supporting edge AI neural networks on STM32 boards is now available. The Board Farm, accessible from the cloud, features a wide range of regularly updated STM32 boards, enabling data scientists and developers to remotely measure the real-world performance of optimized models.

Didier Pellegrin, VP AI Forecasting and Strategy at Schneider Electric, said: “Model Zoo, the STM32Cube.AI online interface, and the remote benchmarking capability for STM32 boards will now allow data scientists with any hardware knowledge to easily evaluate the applicability of AI models to STM32 microcontrollers. Moreover, being able to test their models on multiple STM32 microcontrollers with just a few clicks can help design advanced features with embedded AI processing in mind early in the design process.”

“The STM32Cube.AI developer cloud helps developers validate their approach to developing AI-enabled products in a very short time, and the board farm helps them verify that these models run on a microcontroller,” said Serge Robin, engineer for microcontrollers and digital components at Somfy. “The ability to perform remote benchmarking on different STM32 boards helps designers choose the most suitable STM32.”