STM32Cube.AI

Free tool for Edge AI developers

STM32Cube.AI allows you to optimize and deploy trained Neural Network models from the most popular AI frameworks on any STM32 microcontroller.
The tool is available via a graphical interface in the STM32CubeMX environment as well as in command line. This tool is now also available online in the STM32Cube.AI Developer Cloud.

Download STM32Cube.AI

New in version 8.0

Introducing the support of ONNX Tensor-oriented file format (QDQ)
Update of supported AI framework:
– TensorFlow 2.11
– Keras.io 2.11
– ONNX Runtime 1.12.1

STMicroelectronics STM32Cube.AI solution overview

From Neural Networks to STM32 optimized code

Identify the right STM32 MCU for your project and generate the suitable code from your trained Neural Network model

Load NN model

Analyse NN model

Validate

Optimize

Generate code

Select your MCU and load your trained model from your favorite AI framework: Tensorflow, Pytorch, ONNX, Scikit-Learn.. STM32Cube.AI supports FLOAT32 or quantized INT8 weight formats (input file formats: .tflite, .h5, and .onnx).

The model analysis provides access to a complete set of information about your model, such as the number of parameters, MACC/layer complexity, and detailed RAM and flash size requirements.

Validate your model with a data set or random values to check that the generated C-code matches with the original trained model supplied. Validation options can be performed on the desktop computer or on the STM32 board connected to the computer.

Optimize your model by managing memory usage by layer and choosing the right balance between internal and external memory resources.

Generate the optimized C-code of your AI inference model.

Make the most of your STM32 microcontroller

By optimizing the memory use and the inference time of your AI models, STM32Cube.AI ensures they can easily run on microcontrollers.
STM32Cube.AI is the most efficient free Neural Network code generator for MCU!

Up to

20 %

space freed-up in FLASH and RAM*

Up to

60 %

faster inference time*

* versus TensorFlow Lite for microcontroller

STM32 model zoo – Find the best edge AI model

The STM32 AI model zoo is a collection of pre-trained machine learning models that are optimized to run on STM32 microcontrollers. Available on GitHub, this is a valuable resource for anyone looking to add AI capabilities to their STM32-based projects.

– A large collection of application-oriented models ready for re-training
– Scripts to easily retrain any model from user datasets
– Application code examples automatically generated from user AI model

Go to the STM32 model zoo

Boost your application
with STM32Cube.AI

Using AI opens the door to new application possibilities! Discover inspiring real-world examples and use our resources to create your own application.

Browse use-cases