Edge AI

The STM32N6 has been officially launched yesterday at the STM32 Summit!      The STM32N6 is based on the Arm® Cortex®-M55 running at 800 MHz. It is the first STM32 MCU to embed the ST Neural-ART accelerator™, an in-house developed neural processing unit (NPU) engineered for power-efficient edge AI applications.   A dedicated computer vision pipeline with a MIPI CSI-2 interface and image signal processing (ISP) ensures compatibility with a wide range of cameras. The STM32N6 also features an H264 hardware encoder and the NeoChrom™ Accelerator for graphics, making it suitable for feature-rich products.   4.2 Mbytes of contiguous embedded RAM are also offered, ideal for neural networks or graphics applications, complemented by high-speed external memory interfaces (hexa-SPI, OCTOSPI, FMC).   The full AI ecosystem is out as well! ST Edge AI Core v2.0: introducing the STM32N6 with Neural Art, and also the Stellar P ST Edge AI Developer Cloud sup

Hi,We are trying a very simple deep neural network model with 20 inputs and 3 outputs and want to generate inferences in real-time. We are following the instructions from the document "Getting started with X-CUBE-AI Expansion Package for Artificial Intelligence (AI)" numbered UM2526 with X Cube AI 7.1.0 with STM32CubeMX.Has anyone encountered the same error?Thanks in advance.

I am currently using the STM32 Edge AI Developer Cloud to convert my ONNX model, which I built using PyTorch.The ONNX model is attached for your reference.I would greatly appreciate your guidance on this matter, as it is blocking my research.Thank you for your assistance.  

Dear Sir/Madam,I have a trained CNN model converted to .TFlite file. The file has been analysed and validated on PC and Target. All are correct. Then an application template was developed. When I tried to run initiate CNN model with generated function " static int ai_boostrap(ai_handle *act_addr)", there is issue in this layer:ai_bool cnn1_configure_activations(ai_network* net_ctx, constai_network_params* params) (Can not go deep anymore).  Always return false.  From ai_cnn1_init() returned an error type 48 and error code 19 as shown in the screenshot attached.Would you please help me sorting out this issue? QSAppendix:AI_DECLARE_STATICai_bool cnn1_configure_activations(ai_network* net_ctx, const ai_network_params* params){AI_ASSERT(net_ctx) if (ai_platform_get_activations_map(g_cnn1_activations_map, 1, params)) {/* Updating activations (byte) offsets */ serving_default_Combined_Input0_output_array.data = AI_PTR(g_cnn1_activations_map[0] + 632);serving_default_

..

Hi, When ST will release more information about the STM32N6? Will it use the new cortex-m55? Can it be used in motor control applications?

Hello everyone, I am trying to use the ST Edge AI Developer Tool and I am encountering the following error when I try to quantize my model (a RNN with an input layer, 3 GRU layers and a Fully Connected layer):    2024-12-11 07:25:35.451880: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`. 2024-12-11 07:25:35.479788: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations. To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags. ONNX Runtime Model Optimization Failed! Consider rerun with option `--skip_optimization'. Traceback (most recent call last): File "/usr/local/lib/pytho

Good afternoon,I want to upload a Machine Learning algorithm on the ISPU part of the LSM6DSO16IS. My alghoritm has to identify is the subject is stationary, walking or running. I used NanoEdgeAI Studio to develop the classification algorithm. By doing this I obtain these files, but don't the .ucf file, which is the file that I have to upload on the ISPU part of the sensor. By uploading this files on the Nucleo, it performs the classification correctly, but I want the classification to be done by the sensor. Which is the easyest way to obtain it starting from the files that I obtained?  

Dear Sir/Madam,I am trying to deploy a trained AI model (CNN+LSTM) to STM32F767. When I tried to analyse the TFlite file with the STM32CubeMX, it stopped and gave the following message: Could you please help me sorting out this issue?Thanks QS  

 /* ============= Copyright (c) 2023, STMicroelectronics All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of the copyright holders nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. *THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.

Hi,I'm working with X-CUBE AI.The code seems working but I have a doubt:status_init = ai_network_create_and_init(&network_handle, act_addr, NULL);Is NULL parameter correct ?Why souldn't I need to pass a pointer also for weigths like 'act_addr' ?If it is required how am I supposed to achieve this point without occcupy other RAM space ? (Note that inference output is actually weird, but it could be correct cause I used mockup input data to feed the network)Please, take a look to code and log below:  ....network_data_params.h.... #ifndef NETWORK_DATA_PARAMS_H #define NETWORK_DATA_PARAMS_H #include "ai_platform.h" /* #define AI_NETWORK_DATA_WEIGHTS_PARAMS \ (AI_HANDLE_PTR(&ai_network_data_weights_params[1])) */ #define AI_NETWORK_DATA_CONFIG (NULL) #define AI_NETWORK_DATA_ACTIVATIONS_SIZES \ { 306432, } #define AI_NETWORK_DATA_ACTIVATIONS_SIZE (306432) #define AI_NETWORK_DATA_ACTIVATIONS_COUNT (1) #define AI_NETWORK_DATA_ACTIVATION_1_SIZE (306432) #define

I am trying to validate a CNN on STM32CubeMX with X-Cube-AI 9.1.0. I provided a datainput.CSV and a dataoutput.csv files to  "Validation inputs:  & Validation outputs: " respectively. When ran Validate on desktop or Validate on target, they all failed >>  "E204(ValParamError): number of data files is not enough, 4 are expected."Could you please help me sort out this problem?Do data files in .csv  correct?Only have 2 boxes for 2 files providing, why expected 4? QS 

First , the github project is thishttps://github.com/Xiaobin-Rong/gtcrn/tree/main   It has converted to onnx and tensorRT for ppl to usebut when i load the onnx model into stm32 for optimize/import (setting , compression : none , optimization : balance)it just jumps out an error X-cube-AI version is 9.1.0     ST Edge AI Core v1.0.0-19894 INTERNAL ERROR: Trying to remove node_339 which has multiple inputs     is it because the multi input and output network limitation or something else?and if i success to import this modelhow to deal with the multi dimensional  tensor arrays that the github author demo ?   if you want to test on the modelthere is the link for the onnx and tensorRT modelhttps://github.com/Xiaobin-Rong/TRT-SE/tree/main/online/models   Thanks in advance     Uveuvenouve

Hi everyone,   I have a STWIN.BOX (STEVAL-STWINBX1) and I have played around with the code examples in Python provided on the tfp-sns-datalog2 repository. The idea is to use this device to provide readings to a Raspberry Pi board and make some decisions based on the data. I am not an expert in Python, but I coded a simple script to read the sensors using the repository, getting the hsdatalog_cli.py as an example, and using this data as input to the Raspberry Pi board. The thing is, I can read the sensors only if I start logging (using the start_log method provided by the script hsdatalog_cli.py), and since the log has started, it will generate .dat files. So, every time I run the algorithm inside the Raspberry Pi, new .dat files are generated, and the file sizes depend on how long the board is running the algorithm. I have tried to get data from the sensors without starting logging, but I have failed. Is there any way of getting data from the sensors without generating the dat fil

I am encountering an issue when deploying a classification model generated with NanoEdge AI Studio 4.6.0 to my STM32H7 development board. Below are the details:Issue: When calling neai_classification_init(knowledge), the function returns NEAI_COMPILER_ERROR. Debugging further, I found the following comment in the generated NanoEdgeAI.h file:  enum neai_state { NEAI_OK = 0, NEAI_INIT_FCT_NOT_CALLED = 123, NEAI_BOARD_ERROR, NEAI_KNOWLEDGE_BUFFER_ERROR, NEAI_NOT_ENOUGH_CALL_TO_LEARNING, //This is a fail-safe to prevent users from learning one or even no signals. NEAI_MINIMAL_RECOMMENDED_LEARNING_DONE, NEAI_UNKNOWN_ERROR, /* Error due to a potential compatibility issue affecting user of the ARMCC toolchain. * Contact our support for further details or consider using a different toolchain */ NEAI_COMPILER_ERROR };  Setup:Toolchain: Keil MDK 5.37 with ARMCC (AC5) compiler.Initialization Code: Generated with STM32CubeMX.NanoEdge Versions:Models generated with N

When I select the library in the [Validation] tab of NANOEDGE AI STUDIO and run [Serial emulator], I get the error message "No data are being received. Please ensure the correct COM port is selected."However, I am able to obtain data if I select the same COM port using TeraTerm (terminal emulator).Incidentally, the same error occurs whether the baud rate is 9600 or 115200.The board I am using is B-L475E-IOT01A2.Does anyone have any information on the cause of this?

Hi community,I'm working on deploying an object detection project using Edge Impulse on STM32H747i-DISCO with CAMS OMV. I successfully deployed the TFLite model, but I would like to add camera and board configuration to detect objects in real time. How can I integrate this into the project?Thanks in advance for your help!

Hi, I am getting this error in the Colab note book. please see attached.  Any idea what is wrong?  Also, it seems this is the only Colab note book that could work. any place for a prject that will work with the pre-trained model?  Thank you.     

          Is my camera module working fine (all 3 LEDs light up)? I am working with the model zoo image classification project, and facing issue.I have 

Hi Community,Is there a pre-trained model for object detection, other than for people counting?

Dear Sir or Madam,I'm trying to import a ONNX network with X-Cube-AI(9.1)/ST Edge AI Cloud which was generated and exported using Matlab/Simulink. When trying to work with the model I always receive the following error:[... ST Edge AI Core v1.0.0-19894 INTERNAL ERROR: Your model ir_version is higher than the checker's....] - same error in X-Cube-AI(9.1) and ST Edge AI Cloud.The intermediate representation version of the Matlab/Simulink generated ONNX Model is 9 and I choose OpsetVersion to be 13 (recommendation somewhere in the docs). Attached is an example of the file I would like to import.Now it would be interesting for me to know which intermediate representation version is required by X-Cube-AI(9.1)/ST Edge AI Cloud? Then I can convert the onnx file to the correct ir version.Which ONNX version is the basis for X-Cube-AI(9.1)/ST Edge AI Cloud? Have a nice day and best regards,MoCo STM32Cu

Hi Community, I’m working on deploying my custom tiny ISSIMO-V8 model for object detection onto the STM32H747I-DISCO board with the OMV CAMS camera. Can I use the FP-AI-Vision 1 project, which is based on object detection for people counting, and modify it to adapt to my model (detecting kitchen utensils and multiple objects)? Thank you in advance for your guidance! Best regards,

Hi,I'm facing a weird problem: I correctly deployed a network using x-cube-ai but now I have to add a new one.The configuration option does not open, when I click on x-cube-ai (see screenshot attacched) it stucks on waiting puntator...How  can I fix it ? I tried restarting STM32CUBE IDE and also restarting PC, problem persists...IDE version 6.12.1Regards

Hello,I am interested in implementing a static version of the NanoEdge AI Anomaly Detection for the ISM330ISNTR. That means I would like to us the trained knowledge from NanoEdge AI Studio for the library and not have to run a learning cycle after each power up, because my application requires many short uses, where a learning sequence each time would be very hindering. Is this at all possible?Many thanks in advance :)

Hi everyone,I'm trying to integrate st_yolo_lc_v1_192 pretrained model to my project on STM32H750 MCU.I'm new to the framework and AI tech.I see the output shape is 12x12x30 which I know the 12x12 means there are 12x12 grids. What I don't know is the '30', does that mean 5 * [x, y, w, h, confidence, class] ? and do I need to apply an activation function to these values? Thanks!