Edge AI

With the new X-Cube-AI Version 5.2.0, when I try to validate on desktop/validate on target, the metric for accuracy says "n/a" in the output report, with a note "ACC metric is not computed ("--classifier" option can be used to force it)". Does anyone know where I can access this "--classifier" option or how to get the accuracy of the model again?It seems in the new version the main metric is l2r loss, however, it would be very helpful to get the accuracy metric again.Thanks in advance for any help.

Hello,When I create a new project in CubeMX for the STM32H747I-DISCO board, I cannot see the X-CUBE-AI package when I click on "Additional Software". Is it because this board has 2 cores Cortex-M7 and M4? Somehow we need to specify on which core the network should run?In the user manual of FP-AI-VISION1 (UM2611: https://my.st.com/content/ccc/resource/technical/document/user_manual/group1/a0/53/9b/b7/f9/b5/4a/6f/DM00630755/files/DM00630755.pdf/jcr:content/translations/en.DM00630755.pdf), it is written that:"For users having selected dual-core MCUs for their application but running it on the Cortex®-M7 core only: when dual-core MCUs like the STM32H747 are not supported by the STM32CubeMX tool, they must select a singl

HiI am following this material to start using STM32 for AI. I have got to the data validation step on page 29 of the document. It works fine while validating using random numbers.But while trying to validate using the dataset, I get the error: TOOL ERROR: cannot reshape array of size 1317858 into shape (219643,90,3,1)Could I please know what should be the validation dataset?

Hi all,Currently using the STM32H747I-DISCO with camera to explore using computer vision for people detection. I have already successfully used one of the example NN (found in the FP-VISION-AI1) and now look to use the X-Cube-AI to integrate further tensorflowlite NNs.I'm working off of the Artificial Intelligence (AI) and computer vision function pack​ for STM32H7 microcontrollers - User manual and it states:As such I have loaded CubeMX for the STM32H743ZITx to allow Cube-AI functionality, however, when loading various official tensorflowlite files, I always get the same error message when analysing the model:The .tflite files can be found at: https://tfhub.dev/s?deployment-format=lite&mo

I am playing with my STWIN evaluation kit and I am trying to understand the final step of configuration of the Machine Learning Core in the sensor. Are the configuration registers on the sensor flashed like that of an MCU when you load the ucf file? Is the sensor configured every time by the MCU upon reset?I ask because I am trying to use an MLC sensor in my design where it would be interfaced to am MCU where you program in Micropython. It would take some time to write equivalent drivers.

I have multiple .tflite models that I want to implement in the same project? Is that doable? Also, can I validate data programmatically?Thank you in advanceHanya

I have downloaded x-cube.ai software for converting PyTorch model for MCU. I am unable to install x-cube. After extracting zip file, i don't see any option to install. I see a file after extraction "stm32ai. But unable to run it.Please help me regarding installation of this software.

I am a SW engineer, so I ordered HW devp. to others.They are using STM32F4 board to gather optical sensor data and display prediction(sensor requires over 200 KHz reliable clock freq. at minimum).For prediction data will be transfered to phone via bluetooth and transfered again to cloud server and prediction is transfered in reverse-order.The architecture is pretty complex and comm. overhead is serious, so I am considering to use linux-based system to gather data and predict in one board.Training would be done on desktop and produces keras or tf model, so I need only runtime prediction environment on board.Problem is prediction contains data pre-processing and it needs mainly numpy and scipy library.The other thing is, gathering data would be written in C, and prediction code in python.

Hello,I am using stm32l476, i am trying to run an AI validation project.i replaced the UART with USB (CDC profile)but, when trying to run "validation on target" from the CubeMX, i get this message -> LOAD ERROR: STM32 - no connected board(s), invalid firmware or the board should be re-started.i saw in my debugger that 5 messages received , but nothing was sent to host.what do you think could be the problem?Avi

Hi,I am using stm32l476, i am trying to run an AI validation project.i replaced the UART with USB (CDC profile)but, when trying to run "validation on target" from the CubeMX, i get this message ->"unable to open the port".In debug mode , i can see that 4 messages has arrived and has been replied.Attached, my project with memory snapsohot which shows the arrived messages.What do you think can be the problem, its hard to debug becuase the protocol itself is not described.Avi

Hi,I'm working on lsm6dsox imu for machine learning application. MCL configs worked well on the Unico GUI. Then i passed CubeMx, made requried configurations, took the updated libraries from github(https://github.com/STMicroelectronics/STMems_Standard_C_drivers/tree/master/lsm6dsox_STdC/driver), took the example application from https://github.com/STMicroelectronics/STMems_Standard_C_drivers/blob/master/lsm6dsox_STdC/example/lsm6dsox_mlc.c but mlc feature didn't work. All the other features worked well( read_data_poling, free_fall exc.) What was that case, where is the problem?Thanks.

Hello, Hope you are staying safe&well. I wanted to ask if I can train the model with new data on the STM32 chip rather on python? Thank you Hanya

I confronted an error "TOOL ERROR:  (1, 4, 12, 10) is not a permutation for (2, 0, 1)"It seems there's some issues with handling the model structure. But I can't figure it out.The model perfectly works on python and the structure is not so different from models which had been successfully imported to the my board (NUCLEO-L452RE)The tool version is X-CUBE-AI 5.1.2 and I'm using the STM32CubeMX GUI.The model has the structure shown below (checked by model.summary())_________________________________________________________________Layer (type) Output Shape Param # =================================================================input_8 (InputLayer) [(None, 36, 100, 1)] 0 _________________________________________________________________conv2d_16 (Conv2D) (None, 36, 100, 5) 130 ______________________________________________

I have linux server WITHOUT GUIs. How can I install CubeAI-CLI on that. The idea is to generate files form Keras model with a python script.

Hi,I am using STM32F769I-DISCO. I want to make a system that can recognize different sounds with an already built AI model.I found a document that has an MCU process.I am new to these things so I hope to get some information about:How to stream data from MEMS to audio_preprocessing_library?How to use Audio_Preprocessing_Library? I could not find any example for that.

Hello, When I upload my keras model, I get this error. But the point is I need Keras 2.4, Is there a newer version of X-CUBE-AI that supports it? Thank you!

Hello Team,Which specific ST board to used for AI Appplications or AI product development.I have stm32h7b3-dk or Nucleo-F401RE board , which boards supports running a Machine learning model,Please do let me know.

Trying to start with STM32 and AI. Ubuntu 19.04 STM32Cube.MX 5.3.0 with STM32Cube.AI installed. ,STM32CubeMX IDE 1.0.1. Using an STM32F411CEU6 (8x compression). Following this example"Getting Started with STM32Cube.AI" but using the STM32 Toolchain (nt everyone bought IAR) got the following build errors in STM32Cube IDE.Update #1. Added FREERTOS and in SYS selected Timebase TM1. The CMSIS and tick error are gone but now there are errors as shown in the attached txt file. It seems some constants are not declared Both with FREERTOS CMSIS_v1 and CMSIS_V2. What is missing ?

STM32 AI: Not able to run the example Apps in ST BLE Sensor App, Tried to play the Apps but couldn't run any apps.  Have attached the picture of the example Apps which I couldn;t play as per the ST BLE Sensor App.

I receive the following error when I try to load a model I created with Keras: INVALID MODEL: Couldn't load Keras model, error: Model saved with Keras 2.4.0 but <= 2.3.1 is supportedAs you can see in attached screenshot I'm working with a python virtual environment where I installed Keras 2.3.0, so I am not able to figure out where the problem is. Thanks for the help !

the model I have bulit:​the config of my project:only use cubemx.ai for Application Template，without any other config in my project

I am using CubeAI to load a model from Target Selection. However this model was trained using google Colab. So when I load it and Analyze it I got the error: Unknown Initializer GlorotUniform .Searching through the web the solution provided is not when building the model but when losing the model but since I don't know how to modify how CubeAI loads the model I am asking for some help here

Hello, I am a student studying in Korea.I am doing machine learning using LSM6DSOX sensor and UNICO GUI program.'https://www.youtube.com/watch?v=J9PRrM2YOU'Like the experiment in this youtube video(which is uploaded by your company)We've succeeded in getting real-time machine learning results through the UNICO GUI program, but we don't know how to transfer these results to other devices using Bluetooth communication.Is there any way to check the value of the 'decision tree results' of the UNICO GUI program in other devices?And I'd like to ask if the UNICO GUI program can also run on raspberry pie.Thank you for reading, and thank you for inventing such a great device.

Hi,We are working to an ANN development on TensorFlow, this ANN works on computer and not on MCU.During implementation throught CubeMX, an estimation of RAM and Flash permit to select an appropriate MCU.My question is, ANN implementation is only available if we use internal RAM and Flash or we can run ANN on external RAM ?BR,

Here i only found years old https://community.st.com/s/question/0D50X00009XkZkuSAFI have idea in application detect some cocncrete texts example "HEY ALEXA"and need good probability of detection in ADC received voice.Exist any new examples? Info?