ST AIoT Craft a new web-based tool, simplifies the development and provisioning of node-to-cloud AIoT (Artificial Intelligence of Things) projects using the machine-learning core (MLC) of ST’s smart MEMS sensors.
The MLC is unique to ST’s MEMS portfolio and enables decision-tree learning models to run directly within the sensor. Operating autonomously without host-system involvement, the MLC ensures low latency, consumes minimal power, and efficiently handles tasks requiring AI capabilities, such as classification and pattern detection.
Streamlined Development with Robust Security
ST AIoT Craft integrates all the steps necessary to develop and provision IoT projects leveraging the MLC for in-sensor AI. It offers a secure, user-friendly approach with robust cybersecurity measures for cloud data. As a web-based tool, it eliminates the need for installation, allowing users to access it conveniently online. This not only saves time but also facilitates seamless collaboration between team members, such as AI specialists and embedded software engineers.
Key Features and AutoML Integration
The tool includes an AutoML function, which simplifies the creation of decision-tree models by automatically selecting optimal attributes, filters, and window sizes for sensor datasets. It also trains the decision tree to run on the MLC and generates a configuration file to deploy the trained model. This makes it an excellent entry point for beginners, providing an intuitive introduction to ST’s smart sensors while streamlining the development of AI applications.
For provisioning IoT projects, the tool leverages the Data Sufficiency Module (DSM), which intelligently filters data points for transmission to the cloud. This optimises communication, reduces power consumption, and simplifies future retraining processes.
Ready-to-Use Examples
ST AIoT Craft offers pre-built examples to demonstrate the development of decision-tree-based IoT sensor-to-cloud solutions. These examples include fan-coil monitoring, asset tracking, human activity recognition, and head-gesture detection. They are pre-configured to run on ST reference IoT boards, such as SensorTile.box Pro, STWIN, and STWIN.box enabling users to quickly evaluate the solutions. Users can also customise these examples by incorporating their data or enhancing the provided datasets, accelerating their project timelines.
Seamless Integration with ST Edge AI Suite
ST AIoT Craft is part of the ST Edge AI Suite repository, which contains all the software tools, examples, and models needed for developing machine-learning algorithms for deployment on ST edge-AI devices. These include STM32 microcontrollers (MCUs), Stellar MCUs, and MEMS sensors equipped with the MLC or the intelligent sensor processing unit (ISPU).