Getting Started with STM32N6570-DK AI From Scratch

Description

This application note provides a complete, reproducible workflow for building, deploying, debugging, and programming an AI application on the NUCLEO-N657X0-Q from scratch. Using a quantized INT8 neural network that approximates the sine function, it covers the required software tools and tested versions, model creation and conversion to TensorFlow Lite, project generation in STM32Cube AI Studio, source-code integration in STM32CubeIDE, NPU execution verification, signed FSBL and application binaries, debugger configuration, troubleshooting, and final board programming on Linux and Windows. The document emphasizes practical configuration details, known version constraints, reproducible file paths and commands, and validation of the resulting inference on the target hardware.

Package Summary

Title Getting Started with NUCLEO-N657X0-Q AI From Scratch
Filenames NUCLEO-N657X0-Q-AI-from-the-scratch-v01.zip
NUCLEO-N657X0-Q-AI-from-the-scratch-v01.pdf
License See application note for licensing conditions.
Package content ZIP archive with all files as described in application note.
Size ZIP file: 4 055 477 Bytes
PDF file: 3 148 311 Bytes
MD5 Finger print ZIP: 1fcf62257f9d3c2261bd7cf08e4de265
Installation notes See application note
Revision History V01 - 08.09.2026
V00 - 31.08.2026 PDF ZIP

Result Category

Project number Project acronym Year RIV category Comment
9A25004 NeAIxt 2026 Gfunk Functional sample

Contact Person

Please, do not hesitate to contact Lukas Kohout to obtain more information.