X-CUBE-AI

Data brief

Artificial intelligence (AI) software expansion for STM32Cube

Optimize and validate Generate optimized

Load an own trained neural network model
the NN model code for STM32

or pick one from STM32 model zoo

(AI models library)

STM32Cube.AI for desktop .c .a

STM32Cube Command-line
ecosystem interface STM32 C code
of the optimized model
and AI runtime (.a)
Neural network model converter
for STM32 MCU
via . via .

DT60962V1
Graph Memory
Quantizer
optimizer optimizer

Product status link

X-CUBE-AI

Features
• Generation of an STM32-optimized library from pretrained neural network and classical machine learning models
• Native support for various deep learning frameworks such as Keras and TensorFlow™ Lite, and support for all frameworks
that can export to the ONNX standard format such as PyTorch™, MATLAB®, and more
• Support for various built-in scikit-learn models such as isolation forest, support vector machine (SVM), K-means, and
more
• Support for 8-bit quantized neural network format (TensorFlow™ Lite and ONNX Tensor-oriented QDQ)
• Support for deeply quantized neural networks (down to 1-bit) from QKeras and Larq
• Relocatable option enabling standalone model update during product lifecycle by creating a model binary code separated
from the application code
• Possibility to use larger networks by storing weights in external flash memory and activation buffers in external RAM
• Easy portability across different STM32 microcontroller series through STM32Cube integration
• With a TensorFlow™ Lite neural network, code generation using either the STM32Cube.AI runtime or TensorFlow™ Lite
for Microcontrollers runtime
• Free-of-charge, user-friendly license terms

DB3788 - Rev 10 - February 2023 www.st.com

For further information contact your local STMicroelectronics sales office.
 X-CUBE-AI

Description
X-CUBE-AI is an STM32Cube Expansion Package, which is part of the STM32Cube.AI ecosystem. It extends STM32CubeMX
capabilities with automatic conversion of pretrained artificial intelligence algorithms, including neural network and classical
machine learning models. It integrates also a generated optimized library into the user's project.
The easiest way to use X-CUBE-AI is to download it inside the STM32CubeMX tool (version 5.4.0 or newer) as described in the
user manual Getting started with X-CUBE-AI Expansion Package for artificial intelligence (AI) (UM2526).
The X-CUBE-AI Expansion Package offers also several means to validate artificial intelligence algorithms both on a desktop PC
and an STM32. With X-CUBE-AI, it is as well possible to measure performance on STM32 devices without any user handmade
specific C code.

DB3788 - Rev 10 page 2/8

 X-CUBE-AI
General information

1 General information

1.1 Download information

X-CUBE-AI is available for free download from the www.st.com website.

1.2 What is STM32Cube?

STM32Cube is an STMicroelectronics original initiative to improve designer productivity significantly by reducing
development effort, time, and cost. STM32Cube covers the whole STM32 portfolio.
STM32Cube includes:
• A set of user-friendly software development tools to cover project development from conception to
realization, among which are:
– STM32CubeMX, a graphical software configuration tool that allows the automatic generation of C
initialization code using graphical wizards
– STM32CubeIDE, an all-in-one development tool with peripheral configuration, code generation, code
compilation, and debug features
– STM32CubeCLT, an all-in-one command-line development toolset with code compilation, board
programming, and debug features
– STM32CubeProgrammer (STM32CubeProg), a programming tool available in graphical and
command-line versions
– STM32CubeMonitor (STM32CubeMonitor, STM32CubeMonPwr, STM32CubeMonRF,
STM32CubeMonUCPD), powerful monitoring tools to fine-tune the behavior and performance of
STM32 applications in real time
• STM32Cube MCU and MPU Packages, comprehensive embedded-software platforms specific to each
microcontroller and microprocessor series (such as STM32CubeF7 for the STM32F7 Series), which
include:
– STM32Cube hardware abstraction layer (HAL), ensuring maximized portability across the STM32
portfolio
– STM32Cube low-layer APIs, ensuring the best performance and footprints with a high degree of user
control over hardware
– A consistent set of middleware components such as RTOS, USB, TCP/IP, graphics, and FAT file
system
– All embedded software utilities with full sets of peripheral and applicative examples
• STM32Cube Expansion Packages, which contain embedded software components that complement the
functionalities of the STM32Cube MCU and MPU Packages with:
– Middleware extensions and applicative layers
– Examples running on some specific STMicroelectronics development boards

DB3788 - Rev 10 page 3/8

 X-CUBE-AI
How does this package complement STM32Cube?

1.3 How does this package complement STM32Cube?

The X-CUBE-AI Expansion Package extends STM32CubeMX by providing an automatic neural network library
and classical machine learning library generator optimized in computation and memory (RAM and flash) that
converts pretrained artificial intelligence algorithms from most used AI frameworks (such as Keras, TensorFlow™
Lite, scikit-learn, and any model exported in the ONNX format) into a library that is automatically integrated in
the final user's project. The project is automatically set up, ready for compilation and execution on the STM32
microcontroller.
X-CUBE-AI also extends STM32CubeMX by adding, for the project creation, specific MCU and board filtering to
select the right devices that fit specific criteria requirements (such as RAM or flash memory size) for a user's AI
model.
The X-CUBE-AI tool can generate three kinds of projects:
• System performance project running on the STM32 MCU allowing the accurate measurement of the neural
network inference CPU load and memory usage
• Validation project that validates incrementally the results returned by the neural network, stimulated by
either random or user test data, on both desktop PC and STM32 Arm® Cortex®‑M-based MCU embedded
environment
• Application and mixed-precision quantized template project allowing the building of an application including
multi-network support
8-bit quantized networks and binarized neural networks reduce the required flash memory size and improve the
inference time without significant loss on the network accuracy.
The tool also offers a complete flexibility of the generated code, allowing optimal usage of internal and external
memory.
The X-CUBE-AI tool includes a command-line interface for performing all the analysis, generation, validation, and
quantization steps.
Note: Arm is a registered trademark of Arm Limited (or its subsidiaries) in the US and/or elsewhere.

DB3788 - Rev 10 page 4/8

 X-CUBE-AI
License

2 License

X-CUBE-AI is delivered under the Mix Ultimate Liberty+OSS+3rd-party V1 software license agreement
(SLA0048).
The software components provided in this package come with different license schemes as shown in Table 1.

Table 1. Software component license agreements

Software
Copyright License
component

Copyright (c) 2008 Andrew Collette and contributors

http://h5py.alven.org (see note).
h5py BSD-3-Clause
All rights reserved.
Note: refer to http://docs.h5py.org/en/stable/licenses.html.
All contributions by François Chollet:
Copyright (c) 2015 - 2018, François Chollet.
All rights reserved.
All contributions by Google:
Copyright (c) 2015 - 2018, Google, Inc.
All rights reserved.
Keras The MIT License
All contributions by Microsoft:
Copyright (c) 2017 - 2018, Microsoft, Inc.
All rights reserved.
All other contributions:
Copyright (c) 2015 - 2018, the respective contributors.
All rights reserved.
Larq Copyright © 2020 The Larq Authors Apache License 2.0
ONNX Copyright © 2019 ONNX Project Contributors The MIT License

Copyright (c) 2012-2013 Matplotlib Development Team; All Rights Python Software Foundation,
matplotlib
Reserved Version 2(1)
numpy BSD-3-Clause
QKeras Copyright © 2019 The QKeras Authors Apache License 2.0
Copyright (c) 2007–2018 The scikit-learn developers.
scikit-learn
All rights reserved.
Copyright (C) 2011, the scikit-image team
scikit-image BSD-3-Clause
All rights reserved.
Copyright © 2003-2013 SciPy Developers.
scipy
All rights reserved.
six Copyright (c) 2010-2018 Benjamin Peterson The MIT License

tensorflow(2) Copyright 2018 The TensorFlow Authors. All rights reserved. Apache License 2.0

DB3788 - Rev 10 page 5/8

 X-CUBE-AI
License

Software
Copyright License
component

Copyright (c) 2008–2017, Theano Development Team All rights reserved.

Contains code from NumPy, Copyright (c) 2005-2016, NumPy Developers.
All rights reserved.
Contains CnMeM under the same license with this copyright: Copyright (c)
Theano 2015, NVIDIA CORPORATION. All rights reserved. BSD-3-Clause
Contains frozendict code from slezica’s python-frozendict
(https://github.com/slezica/python-frozendict/blob/master/frozendict/
__init__.py),
Copyright (c) 2012 Santiago Lezica. All rights reserved.
Copyright (c) 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009,
Python Software Foundation,
typing 2010, 2011, 2012, 2013, 2014 Python Software Foundation; All Rights
Version 2
Reserved
Jinja2 Copyright (c) 2009 by the Jinja Team BSD-3-Clause
Copyright (C) 2004-2012, NetworkX Developers
Aric Hagberg <[email protected]>
networkx Dan Schult <[email protected]> BSD-3-Clause
Pieter Swart <[email protected]>
All rights reserved.

1. Matplotlib only uses BSD-compatible code, and its license is based on the PSF license.
2. TensorFlow is a trademark of Google Inc.

DB3788 - Rev 10 page 6/8

 X-CUBE-AI

Revision history
Table 2. Document revision history

Date Revision Changes

17-Dec-2018 1 Initial release.

3-Jan-2019 2 Updated Description.

Added the support of TensorFlow™ Lite, quantization of Keras networks, and

19-Jul-2019 3
command-line interface.
Updated Features and How does this package complement STM32Cube?:
11-Oct-2019 4 • Added the support of TensorFlow™ Lite quantized networks
• Added the use of external memories to support larger networks
Added ONNX support:
18-Dec-2019 5 • Updated Features and License
• Updated figures in Detailed description and cover page
Updated Features and How does this package complement STM32Cube? for
10-Jun-2020 6 Deep Learning frameworks.
Updated What is STM32Cube?
Updated the entire document for deprecated toolboxes (Caffe, Lasagne,
ConvNetJs): figures, Features, Description, How does this package
complement STM32Cube? and License.
5-Mar-2021 7
Added code generation using the STM32Cube.AI runtime or TensorFlow™
Lite for Microcontrollers runtime for TensorFlow™ Lite Neural Networks in
Features.
Added the support for open-source models from scikit-learn and the
generation of classical Machine Learning models in Features and How
15-Sep-2021 8
does this package complement STM32Cube? Updated the cover image and
Figure 1.
Added the support of mixed-precision quantization and binarized neural
network (BNN) for STM32 from QKeras and Larq in Features, License, and
20-Jul-2022 9
How does this package complement STM32Cube? Updated the cover image
and removed Figure 1. X-CUBE-AI overview.
Updated the support for 8-bit quantized neural network format in Features.
14-Feb-2023 10
Updated the cover image and What is STM32Cube?

DB3788 - Rev 10 page 7/8

 X-CUBE-AI

IMPORTANT NOTICE – READ CAREFULLY

STMicroelectronics NV and its subsidiaries (“ST”) reserve the right to make changes, corrections, enhancements, modifications, and improvements to ST
products and/or to this document at any time without notice. Purchasers should obtain the latest relevant information on ST products before placing orders. ST
products are sold pursuant to ST’s terms and conditions of sale in place at the time of order acknowledgment.
Purchasers are solely responsible for the choice, selection, and use of ST products and ST assumes no liability for application assistance or the design of
purchasers’ products.
No license, express or implied, to any intellectual property right is granted by ST herein.
Resale of ST products with provisions different from the information set forth herein shall void any warranty granted by ST for such product.
ST and the ST logo are trademarks of ST. For additional information about ST trademarks, refer to www.st.com/trademarks. All other product or service names
are the property of their respective owners.
Information in this document supersedes and replaces information previously supplied in any prior versions of this document.
© 2023 STMicroelectronics – All rights reserved

DB3788 - Rev 10 page 8/8