general
ST Edge AI Documentation ¶
Welcome to the ST Edge AI Ecosystem ¶
Bringing artificial intelligence to the edge is not just about building a model; it is about turning that model into a reliable, efficient, and deployable product. ST’s AI toolchain supports the entire journey, from exploration and prototyping to optimization, validation, and integration on real embedded targets.
The ecosystem offers a clear path from model development to edge deployment. Start from a pre-trained model in the STM32 AI Model Zoo, adapt it with Model Zoo Services (built on ST Edge AI Core), and keep performance, memory footprint, and target constraints in view throughout. With Developer Cloud, you can benchmark and compare models on remote ST hardware early in the process.
The focus shifts to integrating the model into a product. STM32Cube AI Studio converts trained models into optimized C code and integration-ready project templates, while AI Application Packages provide reference assets that simplify system integration. For sensing use cases, NanoEdge AI Studio generates compact, low-power libraries for anomaly detection and classification on STM32.
The ecosystem also spans the wider STM32 range: X-LINUX-AI brings AI to STM32 microprocessors running embedded Linux, and ISP IQTune tunes image pipelines for better visual quality on products using an ST image signal processor (ISP). Reaching beyond STM32, StellarStudioAI brings edge AI to Stellar automotive MCUs, while MEMS Studio develops and validates AI running directly on ST MEMS sensors.
Together, these tools bridge AI innovation and embedded product realization. They give ML engineers the means to prepare and validate models for the edge, and embedded engineers the confidence to integrate them into optimized, scalable, real-world products.
Tools Overview ¶
Graphical tool to convert, optimize, and validate AI models for STM32, generate optimized C code, and review detailed on-target performance reports.
CLI for model analysis, optimization, validation, and code generation across ST targets. Ideal for scripted / automated workflows. Underlies STM32Cube AI Studio, Model Zoo Services, and Developer Cloud.
Embedded AutoML tool that builds anomaly detection and classification models on STM32, optimized for compact, low-power, resource-constrained devices.
Curated collection of pre-trained models and reference examples for common edge AI use cases, providing ready-to-use starting points and benchmarks.
Python services to train, fine-tune, evaluate, quantize, and deploy models, adapting them to ST targets and production workflows.
Cloud-based environment to benchmark, validate, and compare models on real ST hardware — perfect for remote evaluation and rapid experimentation.
Ready-to-use application assets and integration examples for specific AI use cases, helping you accelerate development and system integration.
Complete AI software stack for ST MPU platforms, enabling AI deployment in embedded Linux environments.
Tool to tune image signal processing (ISP) pipelines on ST imaging platforms, optimizing image quality for computer vision applications.
Desktop environment to design, test, and validate AI and signal-processing solutions running directly on ST MEMS sensors.
Integrated environment to develop, optimize, and deploy AI models on Stellar automotive MCUs.