Subscribe to EDOM TECH Newsletter
2025/12/22
STM32 Model Zoo: Build Edge AI Faster
STM32 Model Zoo: Build Edge AI Faster
Making Edge AI Practical on STM32
When developing AI at the edge, it can be said that knowing is half the battle. With newer AI models, even experienced engineers can find optimization increasing complex. On top of that, guaranteeing performance while working within memory constraints is another hurdle to trip over. This is a barrier to entry that plagues many aspiring developers within the field of edge AI.
To address this, STMicroelectronics is working to make edge AI development more manageable across its STM32 ecosystem, lowering the bar of entry and opening up the gate for prototyping and development of embedded AI applications. This is done through the significant expansion for one of the industry’s most comprehensive libraries of AI models for STM32-based embedded vision, audio, and sensing applications, including wearable smart cameras and sensors, security and safety systems, and even robotics.
A Curated Model Library Built for STM32 Developers
The STM32 AI Model Zoo offers a broad catalog of models for diverse applications, providing all the needed resources on GitHub as well as ready-made solutions for many common AI tasks, allowing engineers to quickly identify models that match their use cases. Categorized and organized in a manner developers can easily navigate, they can efficiently compare and select the most suitable options. These also come with demonstrations and tutorials that illustrate how to use the materials provided, including Face Detection and Face Landmarks models in the latest 3.2 release. Essentially, the Model Zoo is a comprehensive toolbox for STM32 developers.
Fine-Tuning Edge AI for Real-World Deployment
For developers looking to take things a step further, ST ensures the toolbox doesn’t end at pre-built models. Those aiming to fine-tune performance, customize behavior, or maximize efficiency can tap into a dedicated suite of tools and services designed to simplify typically time-consuming workflows.
Through ST’s provided scripts, developers can retrain models with their own datasets, apply quantization to fit strict memory budgets, check accuracy, and benchmark inference performance directly on supported STM32 evaluation boards, including options accessible through cloud-hosted testing environments. This removes much of the guesswork around hardware selection and enables faster, more confident experimentation. When a model is ready to deploy, the process is just as streamlined. Custom AI models can be pushed directly onto a board using deployment services and starter application packages that integrate seamlessly with the Model Zoo ecosystem. In practical terms, this means developers can bring their own datasets, tailor models to their exact needs, and dramatically accelerate their path from concept to production-ready products.
-
STM32L476JE - Ultra-low-power MCU
STMicroelectronicsThe STM32L476xx devices are ultra-low-power microcontrollers based on the high-performance Arm® Cortex®-M4 32-bit RISC core operating at a frequency of up to 80 MHz. The Cortex-M4 core features a Floating point unit (FPU) single precision that supports all Arm® single-precision data-processing instructions and data types. It also implements a full set of DSP instructions and a memory protection unit (MPU) which enhances application security.
[Learn More] -
STM32MP2 Microprocessor Series
STMicroelectronicsWith up to dual Arm Cortex®-A35 and Cortex®-M33 cores, the STM32MP2 series offer advanced edge AI capabilities with its NPU accelerator and the flexibility to run AI applications on either the CPU, GPU, or NPU. Additionally, it supports high-end edge computing use cases, such as machine vision, through its multimedia capabilities. This is enabled by the whole ST edge AI ecosystem offer.
[Learn More]
-
STM32N6 Series - High Performance MCU
STMicroelectronicsThe STM32N6 is based on the Arm® Cortex®-M55 running at 800 MHz, the first CPU to introduce Arm Helium vector processing technology, bringing DSP processing capability to a standard CPU. It is also the first STM32 MCU to embed the ST Neural-ART accelerator™, an in-house developed neural processing unit (NPU) engineered for power-efficient edge AI applications.
[Learn More] -
XMC-2400 - Active Micro Cooling Air-Pump-on-a-Chip
xMEMSThe first-ever all-silicon, solid-state active micro cooling chip revolutionizes thermal management for ultra-mobile systems and next-generation AI solutions. The air pump can push the limits with bi-directional and adjustable airflow up to 39 cubic centimeters of air per second with 1,000Pa of back pressure, enabling mobile devices to sustain maximum performance at lower operating temperatures and without processor throttling.
[Learn More]
