Edge AI

Hello everybody, We are currently evaluating the MLC of the LSM6DSOX and have a couple of general questions regarding feature calculation in MEMS Studio.I've included excerpts from the application note for quick reference. Zero-Crossing:"The feature “Zero-crossing” computes the number of times the selected input crosses a certain threshold. This internal threshold is defined as the sum between the average value computed in the previous window (feature “Mean”) and hysteresis defined by the user."Could you please explain how MEMS Studio calculates this feature for the first window, which doesn't have a preceding window for reference? Peak Detector:"The feature “Peak detector” counts the number of peaks (positive and negative) of the selected input in the defined time window.A threshold has to be defined by the user for this feature, and a buffer of three values is considered for the evaluation.If the second value of the three values buffer is higher (or lower) than the oth

Board - B-L4S5I-IOT01A BoardAI application tool - edgeimpulseIDE -  STM32CubeIDE 1.15.11. AI model generated from edge impulse AI platform for B-L4S5I-IOT01A Board. Package downloaded.2. CUBE mx generate code for B-L4S5I-IOT01A Board.3. Edge-Impulse AI Studio package is included in code.4. Compilation fail happens at ei_dsp_handle.h fileAs I made a package from edge impulse for B-L4S5I-IOT01A and started to build its getting errors as in the screenshot. I already gone through blogs and converted the code into cpp using the option available in cube ide to converts c to cpp. Please help me here.Thanks 

how to configure LIS2DUX12 for Machine learning core(MLC)-ARDUINO IDE

example for the step counter function configuration - arduino IDE 

To STM,Subject: NanoEdge AI package not compile with B-L4S5I-IOT01A Board.Board - B-L4S5I-IOT01A BoardAI application tool - NanoEdge AI StudioCUBEMXIDE -  STM32CubeIDE 1.15.11. AI model generate from NanoEdge AI Studio for B-L4S5I-IOT01A Board. Package download for deployment.2. CUBE mx generate code for B-L4S5I-IOT01A Board.3. NanoEdge AI Studio package include in code.4. Code compile for variable from header file of NanoEdge AI.Please see image file - ModelPackInclude2Code5. Compilation fail on function call of NanoEdge AI.Please see image file - Cmplitn fail on edge ai fun cal Please help to include deployment package to B-L4S5I-IOT01A Board code. Reference :[1] https://www.youtube.com/watch?v=yXMUv_C5FGk&t=101s[2] https://stm32ai.st.com/nanoedge-ai/[3] https://wiki.stmicroelectronics.cn/stm32mcu/wiki/STM32StepByStep:Step4_Sensors_usage#Using_sensors_with_B-L475E-IOT0-

LIS2DUX12-I2C operation-for step counter-how it works

lis2dux12 -mlc implementation .cpp code

How to configure and use machine learning core to develop customized active recognition such as sitting and lying-ardunino ide

I have obtained a model by using Google Colab and translate it through TF Lite. I could apply validate on target and validate on desktop process and I did not get any error. However, once CubeAI generated related .h files and I used them, I got two errors, the errors are attached.Please note that the errors are due to the files that are generated by CubeAI.Thx for your responses.

I need to tune PID coefficients for a set of motor controllers. I’d like to use NanoEdge AI studio, but I’m not sure it supports what I need.  The input to the model is to be a set of four error values and the output is a set of coefficients. NanoEdge accepts UART input for the error values, which is perfect, but I need it to generate coefficients. Instead, I only see examples of it being used for classification tasks. Can it be used for this?   If not, is there another tool?

hello,Now I'm trying to implement ML library created by NanoEdgeAI Studio to our STM32C0 target board.My anomaly detection ML looks work fine at this time, but I wonder that every time after boot the system I have to learn normal case.For our application, sometimes the system will shutdown while the work flows.I also have tried 'Embedded Knowledge' option, but it's not as well so far.So I want to save knowledge after learning while runtime on Real-target, and then it may work fine while the same work flows.When the work to flow changed, then reset (or re-initialize) knowledge and learn normal case in work-setup procedure.  hmm, sounds good.Is this possible?BR, mxo

Hi guys I have a couple of questions. I've been using the microphones on STWINKT1B to record some events for about a month now. I've utilized both the FP-SNS-DATALOG1 and FP-SNS-DATALOG2 firmware for data logging. While the digital mic (imp34dt05) delivers a good signal, there have been instances where the analog mic (imp23absu) produces noisy output with high-pitched tones, either intermittently or throughout the recording. I'm not sure why this is happening, but the interference has certainly compromised the quality of the recording. If anyone's interested, I have attached a recording for review. I'm curious if anyone has any insights into why this might be occurring. My other question relates to the gain settings for these microphones. Is there a way to adjust the gain of the microphone? According to the datasheet of the imp23absu, the output feeds into an op-amp. Is the gain adjustable, or is it fixed? Additionally, according to the JSON configuration file of FP-SNS-DATALOG2, the s

Hello All, I'm currently working with the STM32 H747-DISCO board, but I'm encountering some difficulties in interfacing both the B-CAMS-OMV Camera Module and the LCD. After flashing the camera module, it successfully captures images, and I've also flashed the LCD, which displays the Embedded Wizard GUI. However, only one of them seems to be working at a time. Could you please advise if there are any steps I might be missing? Additionally, I'm seeking guidance on how to interface both the camera and LCD simultaneously. If there are any resources available, I'd appreciate it if you could share them with me.

Can NanoEdge AI only generate.a static library? If so, how do I use it in keil? This is because keil does not use GCC but ARMCC.

I encountered an error exporting the onnx model using FastestDet's model and importing it into CubeAI. I have seen others with a similar error to mine before, but I did not find a solution. What changes do I need to make to deploy this network on my h743.This is my model.This is the model on github：https://github.com/dog-qiuqiu/FastestDetI would appreciate it if you could help me with this problem

Board Used: STM32 Nucleo-L476RGWorkflow: I configured the board, added X-CUBE-AI library in Validation mode, added network in .tflite format with TFLM runtime and generated the code using STMCubeMX. Built the project and flashed the firmware with STMCubeIDE. With the X-CUBE-AI CLI, I validated the tflite model on target.Issue: I cannot understand the profiling results obtained here. Sample output is shown in the picture1. For the duration value, what is meant by 'by sample'?2. And what do the values in brackets represent for the duration? 3. What is meant by HOST duration and how is it different from normal duration? 4. In the TFLM support page on X-CUBE-AI documentation for the 8.1.0 version (screenshot attached), the output snippet provided for AI Validation application does not match the output I get from the CLI in terms of how the duration is displayed, which makes it even harder to understand the output I'm getting. Is the output snippet from a different version of the

Hi, I am trying to generate 9-axis quaternion data using the ISPU function of the LSM6DSO16IS.Hardware environment: nucleo-F401RE + IKS01A3 + LSM6DSO16IS.Software environment: cube ide project generated with MEMS-STUDIO. configuratin file inside ISPU is F:\en.x-cube-ispu\Ispu\ism330is_lsm6dso16is\sensor_fusion_9x\output\sensor_ fusion_9x.ucf.Now I have a couple of queries that I can't understand:1. I understand that the IIC pin of the geomagnetic sensor LIS2MDL should be connected to the MSDA,MSCL pin of the LSM6DSO16IS, and the LSM6DSO16IS takes the initiative to read the data from the LIS2MDL, and then the ISPU generates the 9-axis data with geomagnetic information. However, when I read the production project code, I found that the LIS2MDL data is read/written through the IIC interface of F401 microcontroller, which is mounted on the same iic bus as the LSM6DSO16IS, if so, how can the ISPU get the geomagnetic information to generate the 9-axis data?2. en.x-cube-ispu\Ispu\ism330is_lsm

Hello,I'm sorry that these are in fact many questions but I'm wondering why I can't analyse any of the ONNX models that I've downloaded. I tryied YOLOv8n (12MB), YOLOv6n (18MB), YOLOv5su (36MB), YOLOv4tiny (24MB) and YOLOv2 (62 MB) and none of these ONNX models can pass the analysis step except YOLOv2 (but it is oviously too big for the MCU).I have two main questions.I'm focusing on YOLOv8 since it's the one that I'm fine tuning and I don't understand why I'm getting "INTERNAL ERROR: list index out of range".Vincent Richard from Product Marketing at ST told us that some STM32 users have been able to successfully implement YOLOv2 on STM32. Is it "ST Yolo LC" from model zoo? Otherwise, how can it fit on an STM32?I'm attaching a zip containing all ONNX files and the associated txt files with the errors.Thanks a lot,Camille RAHIData Scientist @ ALTEN

I used the simplest data for anomaly detection, but always failed to generate benchmarks. My setup screenshot with the data set is herelook forward to your reply

内部错误：必须同时设置刻度、零和量化器

How do I convert an .npz log file from gesture evaluation kit application to a .csv file? I tried every possible way but it doesn't convert properly causing errors or giving as output only the labels of the parameters and not the actual values(For example I'm trying to convert the log examples from the handposture dataset)

Hello,I am working on a Human Activity Recognition model with a STMU585CIU6 using Cube AI. In summary, my model takes accelerometer data from my lsm6dsox sensor, and use a frame of 2s of sampling (for ODR 104Hz, it takes 208 samples) as input for the network, which return a prediction among 4 different classes.After struggling for some time to use the generated functions in my main.c, I managed to perform the inference, but not as I expected. I reached a point where I'm not being able to identify and solve the problem.I left my board on the desk to run some inferences, and as you can see in the image, the prediction values are the same for all the classes.I would really appreciate if someone could help me to identify what could be possibly wrong. I am mostly suspicious about the network weights/parameters initialization, but I can't really figure it out. Here is an adapted snippet of my main.c:    ai_u8 activations[AI_NETWORK_DATA_ACTIVATIONS_SIZE]; ai_u8 weights[AI

Hi,I'm working on a university project and the goal is to make a top down people system on the STM32H747i-DISCO. I used the STM32Cube.AI Developer Cloud and used one of the premade models for human detection. However, when I downloaded the project and tried to flash it on to the board I see nothing. No errors, no visuals, no feedback at all. This also occured with the firmware files.Is there anything else I need to do to make this work? I am very new to this.Any help would be appreciated. Thanks,Darrin

 I encountered an issue while using the NanoEdge AI Studio tool for Validation.In step 5, three files need to be uploaded: File Learn, File Regular, and File Abnormal.What data is Flie Name?  

 Could someone provide insights on how to identify the following user actions using a wearable integrated with LSM6DSMTR:Distinguishing between lying down, sitting, and standing.Tracking step count.Detecting if the user is walking on an incline or stairs.Estimating the number of hours of sleep.Additionally, I'm seeking advice on whether it's preferable to process these actions at the edge or to transmit raw data to the cloud for processing. Thanks