STM32Cubex is used in industry

STM32 development system expanded to include AI functions

In order to be able to implement neural networks in end devices and edge nodes, STMicroelectronics is expanding its STM32CubeMX development system. The AI ​​extension STM32Cube.AI contains ready-to-use code examples.

STMicroelectronics has now enhanced its development system for the microcontrollers of the STM32 family, STM32CubeMX, with advanced AI functions. AI uses trained, artificial neural networks to classify data signals from motion and vibration sensors, environmental sensors, microphones and image sensors faster and more efficiently than is possible with conventional signal processing.

With STM32Cube.AI Developers can convert pre-trained neural networks into C code that calls functions in optimized libraries that run on STM32 microcontrollers.

To the extent of STM32Cube.AI includes ready-to-use software function packs that contain code samples for human activity detection and audio scene classification. These code examples can be used directly with the ST SensorTile reference board and the ST BLE Sensor MobileApp. The ST BLE Sensor MobileApp acts as a remote control and display for the SensorTile.

The STM32Cube.AI Extension Pack (part number X-Cube-AI) is available for download in the STM32CubeMX configuration and software code generation system from ST. It currently supports the frameworks Caffe, Keras (with TensorFlow backend), Lasagne and ConvnetJS and IDEs such as those from Keil as well as IAR and System Workbench.

The software function pack FP-AI-SENSING1 contains code examples for seamless motion and audio applications (for recognizing human activity or for classifying audio scenes) based on neural networks. This function pack is based on the SensorTile reference board from ST for recording and labeling the sensor data before the training process. The board can then process the inferences from the optimized neural network.

The comprehensive toolbox consisting of the STM32Cube.AI Mapping tool, application software examples for operation on the compact and battery-powered SensorTile hardware, combined with support from the partner program and the special community, paves the way for the implementation of neural networks on STM32 modules quickly and easily.

Developers receive additional support, including engineering services, from qualified partners as part of the ST Partner Program and from the special STM32 online community for AI and machine learning (ML).

"The new Neural-Network Developer Toolbox from ST makes AI available for microcontroller-based, intelligent devices at the edge and in the nodes as well as for deeply embedded devices in the IoT, in intelligent buildings, in industry and in medical applications," says Claude Dardanne, President of the Microcontrollers and Digital ICs Group at STMicroelectronics.

Related articles

STMicroelectronics GmbH