The first cloud MCU AI developer platform is here! Achieved multiple industry breakthroughs

Today, AI has become an inseparable part of people's daily lives, bringing conceptual impact and changes to every industry and field. It is reshaping the economy, reshaping industries, and pushing the entire human society towards a better future.

Due to the demand for computing power, artificial intelligence technology is currently mainly used in cloud scenarios. Due to limitations in data transmission delays and other factors, cloud-based solutions may not be able to meet some users' needs for data security, system responsiveness, privacy, and local node power consumption. In centralized artificial intelligence solutions, embedded devices (smart speakers, wearable devices, etc.) usually rely on cloud servers to implement artificial intelligence capabilities, while in Deep Edge AI solutions, the embedded devices themselves can run artificial intelligence locally. Algorithms realize real-time environment perception, human-computer interaction, decision-making control and other functions.

Moving the inference process to deep edge computing brings advantages such as system responsiveness, better privacy protection of user information (not all data needs to be transferred to the cloud through multiple systems), reduced connectivity costs and power consumption.

As an expert in the semiconductor industry, ST has always been at the forefront of exploring embedded AI applications. ST has invested a lot of resources in AI, aiming to help developers quickly deploy AI applications on embedded systems based on microcontrollers/microprocessors (STM32 series) and sensors (MEMS, ToF...).

In order to meet the growing market demand for systems based on edge artificial intelligence, ST has previously launched STM32Cube.AI optimized desktop front-end software, allowing developers to optimize and evaluate trained neural network models on the STM32 board, and generate models that can be used on the STM32 board. AI model code that runs efficiently.

STM32Cube.AI is able to interoperate with popular deep learning libraries to convert and apply any artificial neural network to STM32 microcontrollers (MCUs). With STM32Cube.AI, edge IoT devices based on STM32 MCUs can now directly run neural networks, enabling real-time AI calculations at the edge and instant response, thus protecting privacy, reducing network bandwidth usage and consuming a lot of computer power.

Today, ST continues to expand its portfolio of embedded artificial intelligence solutions, providing developers and data experts with an industry-first set of online development tools and services - the STM32Cube.AI cloud developer platform , giving developers the opportunity to use a complete set of industry-focused Online development tools built on the leading STM32 microcontroller (MCU) facilitate software and hardware purchasing decisions, reduce the complexity of edge artificial intelligence technology development, and speed up the launch of new products.

Ricardo De Sa Earp, executive vice president of the General Microcontroller Sub-Products Division at STMicroelectronics, said: “Our goal is to provide high-quality software, hardware and development services to help embedded developers and data experts solve various challenges faster and more efficiently. Easily develop edge AI applications. Today, we launched the world’s first cloud-based MCU AI developer platform. This new tool works closely with our STM32Cube.AI ecosystem to allow developers to remotely test neural network models using STM32 hardware. , saving development effort and costs.”

The newly launched online version of STM32Cube.AI tool - STM32Cube.AI cloud developer platform has achieved many industry firsts:

✦Online graphical user interface : Optimize neural network models for STM32 microcontrollers and generate model C code that runs efficiently on STM32 microcontrollers without prior software installation. Thanks to the industry-proven STM32Cube.AI neural network optimization performance, data scientists and developers can easily and quickly develop edge artificial intelligence projects.

✦ STM32 model library : Contains trainable deep learning models and demonstration application codes to help speed up application project development. Use cases available at the time of tool release include human activity recognition tracking motion sensing, image classification or object detection computer vision, audio classification event detection, and more. These artificial intelligence model libraries are hosted on GitHub and can automatically optimize and generate "Getting Start" software packages that run efficiently on STM32.

✦The world's first online benchmarking service : evaluate the performance of edge AI neural network models on STM32 boards. The cloud circuit board library provides a variety of STM32 circuit boards, and the board library is regularly updated, allowing data scientists and developers to remotely test the actual performance of optimized models using various circuit boards.

This new tool has been tested and evaluated by many embedded development customers around the world, including in China, and has received positive feedback and general praise from them. These customers include Zebra Technologies, Schneider Electric, Husqvarna, Somfy, Lacroix, SIANA Systems, INVT, and more.

Customers have stated that the STM32Cube.AI cloud developer platform allows their data experts and embedded developers to collaborate and share their knowledge on embedded neural networks in an easier way, greatly simplifying the development process and significantly shortening the product development time. Time to market. The platform can also help them quickly evaluate the Flash and SRAM size of the model. More importantly, it can directly evaluate the inference time of the model on the STM32 chip, which can quickly help evaluate system solutions and MCU selection, and select the best performance and cost-effectiveness. Best solution. At the same time, it allows them to consider embedded AI processing in the early design stage, design more advanced functions, develop more innovative products, and maintain market competitiveness.