ST Releases Updated AI Package With More Efficient Machine Learning
STM32Cube.AI Software uses popular Deep Learning training tools widely used by the Artificial Intelligence developer community
STMicroelectronics has expanded the variety of machine-learning techniques available to users of the STM32Cube.AI development environment, giving extra flexibility to solve classification, clustering, and novelty-detection challenges as efficiently as possible.
In addition to enabling development of neural networks for edge inference on STMicroelectronics STM32* microcontrollers (MCUs), the latest STM32Cube.AI release (version 7.0) supports new supervised and semi-supervised methods that work with smaller data sets and fewer CPU cycles. These include isolation forest (iForest) and One Class Support Vector Machine (OC SVM) for novelty detection and K-means and SVM Classifier algorithms for classification which users can now implement without laborious manual coding.
The addition of these classical machine-learning algorithms on top of neural networks helps developers solve their challenges more quickly by enabling fast turnaround time with easy-to-use techniques to convert, validate, and deploy various types of models on STM32 microcontrollers.
STM32Cube.AI lets developers drive machine-learning workloads from the cloud into STM32-based edge devices which reduce latency, saves energy, increases cloud utilization, and safeguards privacy by minimizing data exchanges over the Internet. Now with extra flexibility to choose the most efficient machine-learning techniques for on-device analytics, STM32 MCUs are ideal for use in always-on cases and smart battery-powered applications.
Features of the AI Expansion Pack for the STM32Cube software include:
- Generation of an STM32-optimized library from pre-trained Neural Network models
- Native support for various Deep Learning frameworks such as Keras and TensorFlow™ Lite, and suppport for all frameworks that can export to the ONNX standard format such as PyTorch™, Microsoft® Cognitive Toolkit, MATLAB® and more
- Supports 8-bit quantization of Keras networks and TensorFlow™ Lite quantized networks
- Allows the use of larger networks by storing weights in external Flash memory and activation buffers in external RAM
- Easy portability across different STM32 microcontroller series through STM32Cube integration
- With a TensorFlow™ Lite Neural Network, code generation using either the STM32Cube.AI runtime or TensorFlow™ Lite for Microcontrollers runtime
- Free, user-friendly licence terms
More information on the AI Expansion Pack is available on the ST website
STMicroelectronics AI expansion pack for STM32CubeMX, product page.
*STM32 is a registered and/or unregistered trademark of STMicroelectronics International NV or its affiliates in the EU and/or elsewhere. In particular, STM32 is registered in the US Patent and Trademark Office.
The STMicroelectronics website address is www.st.com
[Reprinted with kind permission from STMicroelectronics - Release Date, 26th July, 2021]