Edge AI

Hi all,I’m currently developing a project on the STM32N6 and am looking for advice on the most robust development workflow. Given the complexity of the N6, I’d like to leverage the automation of CubeMX, but I’ve encountered several hurdles. CubeMX frequently freezes and generates wrong code. Moreover, code generation seems to vary between versions, making it difficult to maintain projects over time or follow existing tutorials (such as https://community.st.com/t5/stm32-mcus/how-to-build-an-ai-application-from-scratch-on-the-nucleo-n657x0/ta-p/828502). Also tutorials recommend to change generated code to make the projects running, which goes against the purpose of automatic code generation.I’ve noticed the ST Edge AI Model Zoo examples appear to use a hybrid approach. It looks like some functions where taken from CubeMX, but the examples itself do not rely on the .ioc file generation.What is the best/recommended way to write software for the STM32N6? Use the firmware package: 

Hi all,I attached my project (CMake using STM32Cube for VSCode) that generates this behavior. I am trying to make a neural network run on the NPU of the STM32N6. I was following this tutorial: https://community.st.com/t5/stm32-mcus/how-to-build-an-ai-application-from-scratch-on-the-nucleo-n657x0/ta-p/828502 , except that I use XIP instead of loading the program to the RAM and use a custom small NN from an .onnx file (also included in the attatched .zip). However, the neural network returns wrong values independent of the inputs and when resetting, the values change.My suspicion is that it has something to do with the cache. However, I am cleaning and invalidating the cache as recommended here: https://community.st.com/t5/edge-ai/the-output-of-a-relative-big-model-is-different-even-if-giving/td-p/811283 (except that I use the function mcu_cache_clean_invalidate_range, which the tutorial recommends)I also save the weights to the flash and not the RAM, which was t

I have a small problem upgrading the STM32N6-GettingStarted-ObjectDetection example project to a newer Edge AI version: I am trying to upgrade the existing yolo person detection net to the latest Edge AI revision (3.0) before exploring custom network.I think I updated all necessary parts as described in the Edge AI version update FAQnew static library (the non- _PIC version)updated the NPU subdirectory, added missing file to the STMCubeIDE projectin main.cchanged the NN name from 'Default' to 'network' to match the generated codereplaced npu_cache_init with npu_cache_enable (at this point, the inference loop was stuck waiting for completion , so..: )implemented npu_cache_enable_clocks_and_reset with the original CACHEAXI enabling codeAt this point, the npu seemingly started but causes this error:Bus interface interrupt BUSIF0 ERR: 0x0 BUSIF1 ERR: 0x3 ATON_STD_IRQHandler()@877: irqs=0xlx assertion "0" failed: file "/home/simon/STM32CubeIDE/workspace_1.19.0/STM32N6-Get

HiWe're looking at an application where it may be helpful for an STM32N6x7 MCU to encode and store video from an ST image sensor then later recover and decode the video for analysis by the Neural-ART Accelerator in the STM32N6x7 - encoding and storing the video in real time then later recovering the video from storage and analysing it using the Neural-ART Accelerator would avoid the need for the analysis to be done in real time and ease power management.Video decoding on the STM32N6 series was last discussed in the community in March 2024 and at that time it wasn't on the roadmap but I thought I'd ask to see if this nay have changed since then?We were wondering if there was any roadmap now for H264 decoding on MCUs or if perhaps a difference approach might be better? For example, storing full quality encoded video and separately storing video downsized to fit the Neural-ART Accelerator as raw rather than as encoded video?The application is an open source wildlife camera for conser

   Hello, I would like to ask about a question regarding “STM32N6570-DK for gesture recognition control applications.”I have downloaded the official example x-cube-n6-ai-hand-landmarks-main and reviewed the DataBrief of the development board. Since the board itself does not contain internal Flash, the file _htmresc/FSBL.PNG in the example clearly indicates two possible solutions:User APP programmed into SRAM (address: 0x3400 0400 ~ 0x3418 0000)User APP programmed into external Flash (address: 0x7008 0000 ~ 0x7FFF FFFF)Following the instructions in README.md, I executed the following steps:~make -j8~ Program the Firmware in the External FlashAfter building the application, it is necessary to add a signature to the binary file:~STM32_SigningTool_CLI -bin build/Project.bin -nk -t ssbl -hv 2.3 -o build/Project_sign.bin~ After obtaining the signed binary file, I manually programmed it to address 0x70100000.(The files ai_fsbl.hex, hand_landmark_data.hex, and palm_det

I've been trying to flash a .tflite model onto my NUCLEO-L433RC-P, using STM32CubeMX and Keil uVision5 initially, and later with STM32CubeIDE. But for some reason I keep getting issues either with "Analyze" or with generating the code. I am on Windows 11. My computer is in the chinese language, but my system local is English (US).​Initially, while using CubeMX/Keil, I was able to successfully analyze my model after downgrading to the 7.1.0 version of Cube-AI. Before that, with the newest version, I get "TOOL ERROR: 'utf-8' codec can't decode byte in position x in position y: invalid start byte". But after analyzing and successfully running the model on desktop, when it comes to generating the code to run on my device, it refuses to generate any .h files nor the middlewares folder. I looked online and couldn't find any solutions, and decided to download CubeIDE as I thought that this might fix the issues. But after downloading CubeIDE, I faced even more issues, as now on top o

I am trying to deploy encoder only transformer model on STM32 by converting pytorch model to onnx format first and then using STM toolchain (STM32 CubeAI). However, I am constantly getting this error when I analyzed the model:TOOL ERROR: cannot reshape array of size 0 into shape () I tried to debug my model but couldn't find any issue in it. Any guidance will be appreciated. Thank you.

Hey Guys,I am currently in the process of running an object detection model on the N6. To make it accelerated on the NPU I quantize it during Training (QAT) in pytorch. I made some adjustments to the exported QAT model in onnx to make it run on the n6 NPU (I compared the original onnx output before and after my transformations and they are the same).When i try to validate my Model on the Desktop (MCU) however i get completely different output for the c model:This is a part of the onnx model in netronThe input is quantized and the output notWhat i observed is when i remove the layer after the Conv->Transpose->Reshape the Output of the C model somehow matches the onnx model:Why is this the case? Is there any problem with the slicing and reshape layer because St has channel last and not first?Thanks for the help,If you need more context let me know

Hi everyone,I am trying to train a large dataset in NanoEdge AI Studio, where each class is around 500 MB in size.Totally 5 class.However, I am facing limitations related to the data structure, memory usage, and training workflow in NanoEdge AI Studio. I would like guidance on:Recommended data structuring or segmentation methods for very large datasetsBest practices to handle or reduce large data sizes (windowing, feature extraction, downsampling, etc.)Any tool limitations or configuration settings to be aware ofPossible alternative approaches if direct training with such large data is not supportedIf anyone has experience training large datasets with NanoEdge AI Studio, your guidance or references would be greatly appreciated.Thank you in advance for your support.Best regards,Vijepradeepan

When using stm32n6 and enabling both GFX and AI models, the runtime continuously reports errors:"Epoch controller error interrupt: EC_IRQ = 0x00000008 Epoch controller opcode counter: 0x00000000 Epoch controller tag: 0x00000000 ATON_STD_IRQHandler()@851: irqs=0xlx Assertion '0' failed: file 'E:/TouchGFXProjects/AR_Glasses/Middlewares/AI_Runtime/Npu/ll_aton/ll_aton_runtime.c', line 855, function: __LL_ATON_RT_IrqErr"What could be the possible cause of this? Is it related to the XSPI2 bus? both GFX and cubeAI utilize XSP2.The model used is the official demo program: st_yolo_x_nano_480_1.0_0.25_3_st_int8.tflite from STM32N6-GettingStarted-ObjectDetection-main.I sincerely hope that st can help me solve this problem. Thanks

Hi everyone,I am trying to deploy a custom ONNX model on the STM32N6 using X-CUBE-AI v10.2.0 (ST Edge AI Core v2.2.0), but the tool crashes immediately during the analysis phase.Setup:Target: STM32N6Tool: X-CUBE-AI 10.2.0Model: GTCRN (from this repository:https://github.com/Xiaobin-Rong/gtcrn)Problem: When I run the analyze command, "stedgeai" fails with an internal Python exception. It seems to be failing on a specific attribute lookup. And here is the output/error log:ST Edge AI Core v2.2.0-20266 2adc00962INTERNAL ERROR: 'range' object has no attribute 'reshape'Has anyone encountered this specific internal error before? It looks like a bug within the converter itself rather than a standard validation error. I would appreciate any guidance on how to fix the model or the export process to avoid this crash.

I am currently working on deploying a quantized Int8 TFLite model (ECG classification) onto the NUCLEO-N657X0-Q board using X-CUBE-AI.The Issue: I have successfully flashed the FSBL, Application, and Model Weights . The system boots up, and the inference loop runs without errors. However, the model always predicts "Class 0" (Normal) , even when I feed it hardcoded input data that corresponds to "Class 1" (Abnormal).The Problem: I performed a deep-dive debugging session by pre-filling the tensor arena (SRAM) with a garbage pattern (0x55) before every inference to check if the NPU is actually writing data.My Findings:Input Data: The input buffer is correctly updated for each sample (verified via printf).Layer-by-Layer Failure:Ep2 (MaxPool) & Ep5 (Logits): The memory at these output addresses remains 0x55 55 55.... This indicates the NPU failed to write to these SRAM addresses.Ep3 (Gemm1): The output is static and does not change with input.Ep4 (Gemm2): Surprisingly, this layer's

Hi!I'm new with STM32CubeIDE and STM32N657X0-Q. I'm working with them because of my bachelor's thesis. My model size es 73KB but 432KB because of activations.The input of my ML is data generated with CMSIS-DSP but I have issues in LL_ATON_RT_RunEpochBlock() line and, I can't see the output.I've followed these guidelines:https://community.st.com/t5/edge-ai/ll-aton-set-user-input-buffer-output-buffer-just-returns-wrong/td-p/842354https://community.st.com/t5/stm32-mcus/how-to-build-an-ai-application-from-scratch-on-the-nucleo-n657x0/ta-p/828502https://community.st.com/t5/edge-ai/problems-with-running-an-ai-model-on-the-nucleo-n657x0-q-board/td-p/818432https://community.st.com/t5/edge-ai/inference-blocking-on-ll-aton-osal-wfe-with-threadx/td-p/868994 This code is not mine, it was used at least to resolve printf issues that I had: https://github.com/YousefDessouki/STM32_N6_NuceloBoardwithNPU/tree/main And I have the following code:LL_ATON_RT_Init_Network(&NN_Instance

Dear ST, I am using Nano Edge AI Studio for Embedded ML teaching at university. Recently, I faced the problem of installing licenses on class computers reporting that “License has reached its activation limits”.I understand that Nano Edge Ai Studio is free but I cannot activate multiple computers in the classroom with my personal free license. Could you please explain how to solve the issue (activate 10 computers in the classroom).Regards,Žilvinas Nakutis

Pricing strategy is the process of making effective plans for fixing the price of the products. Process it accurately and manage your business.A smart pricing strategy is crucial for running a small business. The mentality of a newbie business owner will be on developing the business and attracting clients. Most forget about the crucial factor, which is the pricing. To let you know about the pricing strategy that can be followed by your business, this blog covers the essential details about it and helps you excel in this field.Tap the link to get startedhttps://www.invoicetemple.com 

Great breakdown of PVZ Fusion Neijo. Fan-made mods like this really show how creative the Plants vs Zombies community has become. The fusion system adds a lot more depth to the gameplay and makes experimenting with plant combinations genuinely fun. I especially like how these mods keep the original PvZ feel while adding stronger hybrids and fresh challenges. For anyone interested in learning more about fusion-based PvZ mods, detailed info and updates can be found at https://pvzfusionmod.com/pvz-fusion-neijo/

Hello ST Community,I am currently working on my Final Year Project (PFE) and would appreciate your guidance regarding hardware and system design choices.Project Context (PFE)Objective:Develop a portable Edge AI–based system to detect worker fatigue in an industrial maintenance environment.Planned Technical Stack:STM32MP1 (Cortex-A7 + Cortex-M4)OpenSTLinux / YoctoTensorFlow LitePhysiological sensors: ECG / PPG, IMU, skin temperatureSpecific Questions1. STM32MP1 Board SelectionBetween STM32MP157F-DK2 (~120€) and STM32MP157A-DK1 (~90€), which board would you recommend for this application?Is the price difference justified for the F-DK2 in the context of a student project?Do you have any feedback regarding the stability and reliability of the onboard Wi-Fi/BLE?2. ST-Recommended SensorsAre there STMicroelectronics sensors compatible with the following needs?Heart rate sensor (PPG – ST equivalent to MAX30102?)6-axis IMU (accelerometer + gyroscope)Accurate skin temperature sensorAvailability

1. Problem DescriptionI am running an int8 quantized 1D CNN model on STM32N6570 (NPU / ATON runtime).The full pipeline is:Modbus sensor input → preprocessing → NPU inference → postprocessing (softmax)The issue is that the NPU output (net_out) is identical for every inference, even though:Sensor values are updated periodically via ModbusPreprocessing produces different x_quant values each timeNPU inference runs without errorOutput buffer is properly invalidated after inferenceExample output log (repeats every inference):Despite different input values, the raw int8 output vector is always the same.2. What has been confirmed (important)1) Pipeline connectivity is correctThe full data flow is confirmed to be connected:ModbusTask updates sensor_valsupdate_timeseries_buffer(sensor_vals) is calledPreprocess_Data() runs successfullyx_quant changes every inference ([DEBUG] x_quant CHANGED)x_quant is copied into net_inNPU inference runs (Run_Epoch())Output buffer is invalidatedPostprocess runs o

Hello, I would like to know if I can use FP-AI-Vision to design my own application, for example, to detect trees in a yard. Or is it designed only for the three applications: food recognition, person presence detection, and people counting?

Hi After the latest update, I can't download C Code from Edge Ai Developer Cloud, there are no boards in the list.   fv

Hello. I am currently attempting to execute a MatMul operation on an NPU. I have implemented a simple TFLite file with a MatMul operation, as shown in the image below.When converting the model using ST Edge AI, I observed that Matmul operator is mapped to the MCU instead of the NPU, and the operation is converted into a Convolution. Furthermore, checking the network.c file reveals that it is being converted into a convolution layer with an extremely large stride. https://stm32ai-cs.st.com/assets/embedded-docs/stneuralart_operator_support.html states that the matmul(batch matmul) operator is supported on the ST Neural art Accelerator.How can I resolve this issue? Thank you.

Hello,I recently deployed multiple individual PyTorch models onto my STM32H7A3ZI-Q. My workflow ended up being:1. Convert from .pt -> .onnx -> .tf -> .tflite (quantized)2. Build an X-CUBE-AI project via CubeMX3. Swap models in same project using ST Edge AI Core CLI4. Writing scripts to semi-automate this flowIt works, but setting this up involved quite a few manual steps. Is this the most effective manner to approach this? Curious how others handle thisRegards,Ernest

I’m currently trying with the STM32N6570-DK board using the ST Edge-AI v2.2.0I would like to know whether a lightweight Transformer model (for example, a small Vision Transformer or a minimal Transformer encoder) can be successfully converted and executed on the STM32N6. I checked that batch_matmul, transpose, etc. are supported on the link (https://stm32ai-cs.st.com/assets/embedded-docs/stneuralart_operator_support.html )Currently, after converting the lightweight ViT model to tflite or onnx, I am trying to convert it using stedgeai and stm32cubeide, but only the following error messages appear and are not being convertedCode, model file, and error text: https://github.com/minchoCoin/lightweight_vit These model was successfully analyzed with STM32Cube.AI MCU runtime on STM32CubeIDEFollowing error messages appear when I try to convert these model(tflite and onnx) with STM32Cube.AI Neural-ART runtime in STM32CubeIDE or with stedgeai in command line1. tflitestedgeai genera

HelloI'm considering running an LSTM(RNN-based model) model file, on the STN32N657 NPU.However, I understand that the NPU is CNN-based and therefore cannot run on it.Is there a way to run an LSTM model file on the NPU?I understand that theoretically, it's possible to convert an LSTM to a 1D CNN or TCN.Has anyone converted an LSTM model to a 1D CNN or TCN and tested it on the NPU?Is there an example file available?Thank you. 

Hello, I am working on implementing an object detection application for the STM32N6570-DK, starting from the reference projects:STMicroelectronics/x-cube-n6-ai-people-detection-trackingSTMicroelectronics/STM32N6-GettingStarted-ObjectDetection I'm encountering a blocking issue when implementing the inference pipeline: During debugging, I found that the inference execution blocks indefinitely at LL_ATON_OSAL_WFE() on line 105 of Appli\App\Src\app_nn.c. The code is stuck in the following loop:static void NN_RunInference(void) { LL_ATON_RT_RetValues_t ret; do { ret = LL_ATON_RT_RunEpochBlock(&NN_Instance_od_yolo_x_person); if (ret == LL_ATON_RT_WFE) { LL_ATON_OSAL_WFE(); // <-- Blocks here } } while (ret != LL_ATON_RT_DONE); LL_ATON_RT_