Article Overview

This article introduces NanoEdge AI Studio, a one-click PC development platform for generating and deploying multiple types of machine learning libraries to build edge AI solutions. It supports multiple sensors and optimizes performance on STM32 microcontrollers. This article also lists a variety of STM32-based products supported by DigiKey NanoEdgeAI technology.

About NanoEdge AI Studio

NanoEdgeAIStudio is a new machine learning (ML) technology that provides developers with the privilege to create optimal ML libraries based on minimal data. It is a one-click PC-based development platform that runs on Windows, Linux Ubuntu.

The development platform can generate four types of libraries: anomaly detection library, outlier detection library, classification library, and regression library . These libraries can be combined and linked to create a complete edge AI solution: anomaly or outlier detection to detect problems on the device, classification to identify the source of the problem, and regression to infer information and provide real insights to maintenance teams.

Multiple sensors can be combined in a single library or multiple libraries can be used simultaneously. This topology can respond to various types of input signals such as vibration and pressure, sound, magnetism, time of flight, etc.

Both learning and inference are done directly inside the microcontroller via the NanoEdge ™ AI on-device learning library, which simplifies the edge AI process and significantly reduces development effort, cost and therefore time to market.

All features

• Desktop tool for designing and generating STM32 optimized libraries with small data sets:

1. Anomaly detection libraries : Learn normality and detect defects in real time directly on STM32 microcontrollers
2. One-class classification libraries : perform collection during normal device operation and detect any abnormal pattern deviations
3. N-class classification libraries : classify signals in real time

4. Extrapolation libraries : predict discrete values ​​based on previously unseen patterns in the data

• Supports any type of sensor for various physical quantities: multi-axis acceleration, current, magnetic field, voltage, temperature, sound pressure, etc.
• There are millions of possible algorithms that can be used to find the optimal library in terms of accuracy, confidence, inference time, and memory footprint
• Generates libraries with a very small footprint that run down to the smallest Arm ® Cortex ® M0 microcontrollers

• Integrated tools such as:

1. Sample Finder tool to easily select the correct data rate and the correct data length
2. Data logger generator, ready to log data in just a few clicks
3. Data manipulation tools for datasets
4. ML library benchmark to find the best combination between preprocessing and machine learning models
5. Embedded simulator to test the library performance in real time with a connected STM32 board or test data files
6. Inference time estimates to help users make informed choices for model selection
7. Validation tool to compare libraries provided by NanoEdge ™

STEVAL-STWINKT1B

STEVAL-MKBOXPRO

STEVAL-PROTEUS1 is an evaluation tool designed for temperature and vibration monitoring.

STEVAL-PROTEUS1

B-U585I-IOT02A

The STM32H747I-DISCO Discovery Kit is a complete demonstration and development platform for the STMicroelectronics STM32H747XIH6 microcontroller, designed to simplify user application development.

STM32H747I-DISCO

For more Development Tools related technical content , please check the following:

• Instruction Set Architecture and Its Selection in Embedded Systems
• Introduction to PLC Programming Language
• Solution and demonstration of multi-motor control using the new 32- bit MCU series
• Intel EK-10M08E144 MAX ® 10 Evaluation Kit does not blink
• What is the difference between the Adafruit ESP32 Feather evaluation board 5438 and 5400 ?

"Star" us to avoid missing out on new cases and industry information