Edge AI

Hi,i was trying to update my cubeAI pack but cubeide can't unzip this new version.If I use my old projects and I try to change something in the .ioc file a message appeare on my pc:these IPs still have some not configured or wrong parameter values: [STMicroelectronics.X-CUBE-AI.6.0.0, Verify the Platform Settings Tab ]I was trying to configure my pinout to connect a F401RE to a lcd16x2 QAPASS with an .ioc that I had configured 1 months ago. When I upload my neural network and I try to generate code I have this problem.Now for my project I'm using old ones with pre-uploaded neural network but if I open a new project I still have probems.Thanks

X-Cube AI initialization stucks on ai_network_create() when CRC peripheral is off. I noticed this problem when i tried to use X-Cube AI library inside a C++ class and called ai_network_create() inside a constructor. Class constructor was called before main(), so CRC peripheral was not initialized at that time.   I propose to change ai_network_create() to return an error when CRC fails or is off, instead of stucking in an infinite loop.  

In the above figure what is Basic and template and in the below figure why the other items are disabled

Just received this email: We have had the SensorTileBox SL-MKSBOX1V1 since early this year (and I don't think it was new then) - so what is actually new here?

Can STM32L1 and STM32L0 be used for AI development (is. running inference of neural networks)? I was hoping to use those chips with X-CUBE-AI. Yet, when I read the user manual, those chips were left out of the list of microcontrollers. Why is this so?User Manual: https://www.st.com/resource/en/user_manual/dm00570145-getting-started-with-xcubeai-expansion-package-for-artificial-intelligence-ai-stmicroelectronics.pdfAlso, there is STM32Cube.AI which I assume will work with STM32L1 and STM32L0. Is this correct?

Hello,Device Configuration Tool Console doesn't open in STM32CubeIDE.I clicked "Open Console -> 4 Device Configuration Tool Console". But the console was not opened.It is important to Validation of X-CUBE-AI application.Thank you.

Hello Community team,a new article is now available in the Knowledge articles section:Fan rack condition monitoring with industrial sensor machine learning core --- 1/2Fan rack condition monitoring with industrial sensor machine learning core --- 2/2It describes in all the details how to use the Ma

Hello everyone,​I am trying to deploy ML model to microcontrollers and I have two boards of STM32L4R9I-Discovery​ and STM32H7B3I-Discovery. The experiments and related results are shown below:I use TFLite Micro and STM32Cube.AI runtime to profile my ML model and then compile on two different boards. The hardware resource related to these two boards are provided and STM32H7B3I is a bigger one. However, it's strange that I can successively get the results on smaller board (STM32L4R9I) but build fail on a bigger one (STM32H7B3I) with the following error:.../arm-none-eabi/bin/ld.exe: *****.elf section `.bss' will not fit in region `RAM'​.../arm-none-eabi/bin/l

When the binary `STM32H747I-DISCO_Webcam_V300.bin` is dropped onto the DISCO board the webcam application functions. However, when I build from source the webcam is detected and I can see the resolution options in the `windows camera` app, but the frame is just black ... The code builds with no errors. I am wondering what the difference is between the pre-built binary and the source?

Hi all,I'm starting on AI and so I'm following the tutorial "Getting Started with STM32Cube.AI".  After the first settings I see the message attached. Why?Any Helps?Thanks

I have been following the FP-AI-VISION1: Computer Vision application video tutorial by Thibaut Vercueli and he uses Google Colab and a notebook called "TransferLearning" to train a new Tensor Flow Lite model on custom images, before loading this into Cube.AI I would like to find/use the same Colab Notebook or maybe someone could link me to a tutorial showing a similar approach with the notebook available.

The standard way to generate a header file with MLC register settings is to generate from decision-tree .txt file (exportet from WEKA or python).In our project we want to avoid tool interaction.The register settings (ucf or header file) should be generated by a script (python....).Is there a way without user interaction to create the register settings from (weka) decision trees?regardsHari

I am working on the deployment of the CNN model on an STM32F29 board with 8-bits precision.I am able to generate the validation code, but the problem is that the model data are in u64 format (as illustrated in the image below)please help?thanks,

Following is my data , I am trying to validate on option of 'validate on desktop'. I am getting following errorLoading: /home/censdata/workspace/har/har_ai/har_ai.ioc projectInitializing: STM32F429ZITx [AI:network] Analyzing model [AI:Generator] Command /home/censdata/STM32Cube/Repository/Packs/STMicroelectronics/X-CUBE-AI/3.4.0/Utilities/linux/generatecode --auto -c /tmp/mxAI227727002133516706206695419651859/config.ai [AI:Generator] Python generation ended [AI:network] Analyze and code generation complete on AI model [AI:network] Starting AI validation on desktop with custom dataset : /home/censdata/stm_ai/HAR-CNN-Keras-master/data_compressed/data_3_Axis.txt... [AI:Validator] Command /home/censdata/STM32Cube/Repository/Packs/STMicroelectronics/X-CUBE-AI/3.4.0/Utilities/linux/validate - -p /home/censdata/stm_ai/HAR-CNN-Keras-master/model.h5 -t keras_dump --co

..

Hello,I used the example DCMI_CaptureMode for the STM32F769I-EVAL and the CubeMX generator with CubeAI in order to make an object classifier on board. It worked well with the internal RAM for 28*28*1 pictures but now that I want it to work for 128*128*3 pictures, it needs more RAM, but I can't achieve to use the external SDRAM:The issue is: if I use MX_X_CUBE_AI_Process(); the screen freezes, and if I don't use it, I still have issues calling BSP_LCD_FillRect() => it doesn't show a rectangle whereas on other projects, it did.Question 1: Can we see how much of the external RAM is reserved to CubeAI ? (in order to use the rest of the external RAM for peripherals)Question 2: Do you have an idea on how to deal with my issue ? Anything would be welcome.Here is the way I tried to tackle the issue:-> I used the advanced option to use the external SDRAM for CubeAI, but it comes in conflict with

I'm trying to deploy a machine learning model to an STM32H743ZIT6 through cube-ai extension. The model is in .tflite format. When I try to analyze the model this is the result:The tool states that the MCU has 512KB total available and the model exceeds it but on the datasheet I find 1024KB.What's the reason?

I convert .ucf file to .h by using unico GUI. But I couldn't set the register values via HAL I2C functions. I couldn't get any decision tree result./* ****************************************************************************** * @file LSM6DSOX_configuration.h * @author Sensors Software Solution Team * @brief This file contains the configuration for LSM6DSOX_configuration. * ****************************************************************************** * @attention * * <h2><center>© Copyright (c) 2020 STMicroelectronics. * All rights reserved.</center></h2> * * This software component is licensed by ST under BSD 3-Clause license, * the "License"; You may not use this file except in compliance with the * License. You may obtain a copy of the License at: * opensource.org/licenses/BSD-3-Clause * */   /* Define to prevent recursive inclusion -------------------------------------*/ #ifndef LSM6DSOX_CONFIGURATI

FP-AI-VISION1 has many interesting examples. However, all of them are based on STM32H747I-DISCO. Can anyone provide some instructions and ideas about how to extend the examples in FP-AI-VISION1 to other boards, such as STM32H7B3I or even STM32F7 and STM32L4 series boards?Many thanks for all your help!

This talk at TinyML Summit explains some of the key techniques for optimizing CNN models. The focus in on audio models, but some also apply to other multi-variate time-series and some to image processing.https://www.youtube.com/watch?v=ks5kq1R0aws&t=0s

i will work with :1- STM32F103C8 MCU code as main machine controller2- ESP32-WROOM-32D as communication Processor for Serial, Bluetooth, and WIF

Hi!I watched the video from the link below and was very interested in testing the application, but before purchasing the development boards, I would like to know where the sample codes for this demo can be found.https://www.youtube.com/watch?v=HUeyku7tX7Y&t=27sI searched and didn't find it, could someone tell me where I get these example codes?Thanks!

"ImportError: cannot import name 'get_all_providers' from 'onnxruntime.capi._pybind_state' (C:\Users\Systemname\STM32Cube\Repository\Packs\STMicroelectronics\X-CUBE-AI\6.0.0\Utilities\windows\onnxruntime\capi\_pybind_state.pyc)"Please help me to solve this issue.

Hello,We are trying to convert a pre-trained model using X-CUBE-AI on STM32H747IDISO board. Analyze failed due to RAM usage is exceeding 512KB. When we enable Use External RAM & Use Activation Buffer option in advanced setting, it is still failed. RAM usage is not reduced. Please let us know what other option need to enabled to use external RAM?

Hello , I am trying to use STm32H7 with special camera to do image classification.What AI techniques and what features are used in FP-AI-FACEREC? and if CNN is used, I would like to know what kind of CNN is used for FP-AI-FACEREC.Is it MobileNet or something else?