Getting Started with STM32Cube.AI

X-CUBE-AI is an STM32Cube Expansion Package extending STM32CubeMX capabilities with automatic conversion of pre-trained Neural Network and integration of generated optimized library into the user's project. The X-CUBE-AI Expansion Package offers also several means to validate Neural Networks models both on desktop PC and STM32, as well as measure performance on STM32 devices without user handmade ad hoc C code.

Key Features
- Generation of an STM32-optimized library from pre-trained Neural Network models
- Support various Deep Learning frameworks such as Keras, Caffe, ConvnetJS, and Lasagne
- Easy portability across different STM32 microcontrollers series through STM32Cube™ integration
- Free, user-friendly license terms

[sorrefly]

This is literally awesome. I'm using it to classify EMG response in muscular contraction for prosthesis hand

[fernando.liozzi.]

Hello thank you very much. I did a course on neural networks with keras in EDX, and I can now train, save and load my networks on an STM32F4VE, with compression 8. The verification with my data set works well. And now how do I continue? How do I send my data to the STM32 to make predictions? Thanks

[ttaylor]

@3:10 get error "invalid network".

[SonefromChina]

very good! i was working on tensorflow and cmsis-nn convert for months, thanks for this tool, maybe it can finish my work in days

[embedded]

Hi，the code is successfully generated and built on cubeide，download to the board and it passed the validation on desktop.However it says fatal error Stm32f4xx_hal.h No such file or dictionary when I click the validation on target.What may lead to this happen, please help me.

[linben]

I follow the all instructions but get stuck on analyze step at 5:54. It shows up a "please wait" window and does nothing. The progress bar is even freezed.

[AkshitSharma]

the output of the model shows that it is running with the inputs (10, 90, 3, 1).it is not clear from this output where these inputs came from or how they were generated.

[johnhe]

Actually, I follow the step in the video and face a problem when I validate on desktop. It shows there are 12r errors.

[fernando.liozzi.]

Hi, once I passed the verification on the microcontroller. How does the thing go? How do I enter data by usart, i2c, spi, etc. and that the neural network make the prediction? This only loads verification modules. I want prediction modules. Greetings.

[POLIVERITO]

Hi. I have a STM32F103C8T6 board. Could I do an application of IA? the easiest.... Would you mind helping me please?

[sailfromsurigao]

Does this mean that STM32Cube.AI can be used on most STM32 MCU's for Image Processing, including STM32F407?

[TheSlntOne]

What mcu family is best for autopilot for drones

[alal]

Where can I get the file " har_github_test_vectors.csv" ?

[jsonslim]

Hello! I'm trying to load Keras model but it always fails with "Invalid network" Error. What is might be wrong? I've downloaded model from link on the video

[sammertens]

Hi, I have a STM32L053. Is it not compatible ?

[bhuvaneshs.k]

How to collect real time data not CSV... Using SPI Interface of accelerometer

[wuqiwang]

I creat my own model, the minimum flash is 2.83Mbytes, when i choose compression of4 or 8, it dosn't get much smaller, still around 2.6Mbytes

[karan_rathore]

I am following this video this, i have successfully generated the code, validate it on desktop, build the code and flash using Truestudio it into STM32L476RB board. But when I try to validate it om my board it gives me


error : board not found or invalid firmware or restart the board


I have also tried using different USART port, but no success.


I have aslo flash the code using CubeProgrammer. but still same error on validating.


CubeMX is able locate the port on which board is present but gives same error