Getting Started


Grove Vision AI Module Sensor represents a thumb-sized AI camera, customized sensor which has been already installed ML algorithm for people detection, and other customized models. Being easily deployed and displayed within minutes, it works under ultra-low power model, and provides two ways of singal transmission and multiple onboard modules, all of which make it perfect for getting started with AI-powered camera.

In this wiki, we will show you how to utilize the Grove Vision AI module Sensor connecting seeed studio XIAO Series and Arduino, to detect people, panda, and then display it on the website.


Parameters Description Note
Grove (Grove base for Arduino) 5V Charge and Data Transmission
Double row 7pin socket (seeed duino XIAO) 5V Charge and Data Transmission
USB Type-C 5V Charge and Firmware burn
Communication Mode IIC
Processor Himax HX6537-A 400Mhz DSP (ultra low power consumption)
Camera Sensor OV2640 chip Resolution Ratio 1600*1200
Microphone MSM261D3526H1CPM chip -26dBFs sensitivity
Accelerometer Sensor STLSM6DS3TR-C Sensor 3D accelerometer and 3D gyroscope


  • Easy-to-use AI Camera: Apply Edge Machine Learning algorithm in the camera sensor perfectly for detecting obejcts
  • Plug and Play: Make toilless effort to perform full function and display within minutes
  • Ultra-low Power consumption: Activate while detected objects moving for power saving
  • Compact AI-powered Camera Sensor: The device itself is thumb-sized, vision-based, and suitable for Edge Intelligence
  • Customized Sensor Design: Support custom ML models by users define
  • Two Signal Transmission Supported: Refer to seeed studio XIAO elegant connecting and Grove base for Arduino simple connection
  • Onboard Camera Sensor: Corporate OV2640 chip with 1600*1200 resolution ratio
  • Onboard Microphone: Corporate MSM261D3526H1CPM chip with -26dBFs sensitivity
  • Onboard Accelerometer Sensor: Corporate STLSM6DS3TR-C sensor for 3D accelerometer and 3D gyroscope detected
  • Onboard 32 MB SPI Ultra Low Power Flash
  • One wire for all data output: Represent Grove family simplified connection for only one Grove cable required between the sensor and the single board

Hardware Overview

We assume you might want to be aware of some basic parameters of the product. The following tables present information about the characteristics and pinout of Grove Vision AI Module Sensor.

Characteristic Value Unit
Operating Voltage 5 V
Rate 115200
I2C Interface seeed studio XIAO & Arduino -
Power Supply dual 7-pin connector & Tpye-C -
Downloading & Firmware Burn Interface Type-C -
Dimensions 40*20*13 mm

Pinout Overview

  • 2 —— BL702
  • 3 —— MIC MSM261D3526H1CPM
  • 4 —— 6-axis LSM6DS3TR-C
  • 5 —— SPI Flash
  • 9 —— USB type C
  • 10 —— Dual 7-pin female socket
  • 11 —— Reset button
  • 12 —— BOOT button
  • 13 —— USER button
  • 1 —— HX6537-A
  • 6 —— Camera Connector
  • 7 —— DC-DC Chip
  • 8 —— Grove Connector
  • 14 —— Power Light
  • 15 —— Burning indicator light

Getting Started

We will show you the basic function about the module, and then introduce you the customized way that you can build the ML model of your own. But before we fully apply the module to our projects, it will take us serval steps to get the module ready.



  • Grove Vision AI Module Sensor
  • Windows host PC (Win10 tested)
  • Type-C cable

Arduino Library Overview


If this is your first time using Arduino, we highly recommend you to refer to Getting Started with Arduino.

The code we use here provides serval methods of classification and selectable models, including customized models. The default we select here is object detection method and pre-trained model.

seeed studio will develop more methods and models for the foreseeable future, stay tuned with us.


Before we get started developing a sketch, let's look at the available functions of the library.

  1. if (ai.begin(ALGO_OBJECT_DETECTION, MODEL_EXT_INDEX_1)): —— This is where we can select our methods of classification and the models.


Meanwhile, if you also burn your customized model(firmware) into the module, you can change MODEL_EXT_INDEX_1 to MODEL_EXT_INDEX_2 or 3, 4, as you named it.

  1. object_detection_t data: This is the struct dataset where the result output.

If you change the classification method, you need to change object_detection_t correspondingly to image_classification_t or object_count_t.

  1. uint8_t len = ai.get_result_len() The "len" here means how many human face it detects.

  2. object_detection_t data The data here represent a struct format and the tpye is "object_detection_t" that is pre-defined.

Meanwhile, if you choose other methods of classification, you need to change it to image_classification_t data or object_count_t data.

  1. Serial.print(data.confidence) The struct format can not be directly used to print, it should point to the specific type which is defined in the head file.
typedef struct
    uint16_t x;
    uint16_t y;
    uint16_t w;
    uint16_t h;
    uint8_t confidence;
    uint8_t target;
} object_detection_t;

Hint: The "confidence" we want to print here shows how much "confidence" the camera has to detect the object.

Library Installation

Since we have downloaded the zip Library, open your Arduino IDE, click on Sketch > Include Library > Add .ZIP Library. Choose the zip file you just downloaded,and if the library install correct, you will see Library added to your libraries in the notice window. Which means the library is installed successfully.

Library Upgrading

In the foreseeable future, we will optimize and upgrade the product library for more interesting function. According to the library installation methods provided above, we here introduce you the way to upgrade.

We will update the link when the library is optimized. You can delete the original library folder in your computer's folder, then download the latest version, unzip it and put it in the Arduino IDE library directory. (...\Arduino\libraries. .... is the path you setup Arduino)

seeed studio XIAO / Seeeduino(Arduino UNO) Example

Now that we have our library installed and the firmware burned. we can now run some examples with Grove AI module sensor on seeed studio XIAO BLE Sense and seeeduino V4.2 to see how it behaves.

Step 1 (seeed studio XIAO). Parpare a Type-C cable and connect it to one seeed studio XIAO Series board. Plug it pin by pin into the Grove AI Module and use another Type-C cable to connect the module.

Both Type-C cable should be connected with the PC. In the end, the direction of the Type-C connector on the module should be the same as Type-C connector on the seeed studio XIAO samd21. For instance:

Step 1 (Arduino). Parpare a Grove cable and connect it to one Arduino board. Use another Type-C cable to connect the module.

Step 2. You need to Install an Arduino Software.

Step 3. Launch the Arduino application.

Step 3. The code is Select your development board model and add it to the Arduino IDE.

  • If you want to use seeed studio XIAO SAMD21 for the later routines, please refer to this tutorial to finish adding.

  • If you want to use seeed studio XIAO RP2040 for the later routines, please refer to this tutorial to finish adding.

  • If you want to use seeed studio XIAO BLE for the later routines, please refer to this tutorial to finish adding.

  • If you want to use Seeeduino for the later routines, please refer to this tutorial to finish adding.

  • If you want to use Arduino Uno for the later routines, please refer to this tutorial to finish adding.

Demo 1 Human Detection with seeed studio XIAO BLE Sense / Seeeduino V4.2

In this demo, we will detect human face and count how many people the module detects on both seeed studio XIAO BLE Sense and Seeeduino V4.2 (Arduino UNO). Meanwhile, seeed studio provives a website to display what the module sees.

Additional Materials Required

For seeed studio XIAO BLE Sense

seeed studio XIAO BLE Sense Grove AI Camera
Get ONE Now Get ONE Now

For Seeeduino V4.2

Seeeduino V4.2 Base Shield Grove AI Camera
enter image description here enter image description here
Get ONE Now Get ONE Now Get ONE Now

Code Uploaded

#include "Seeed_Arduino_GroveAI.h"
#include <Wire.h>

GroveAI ai(Wire);
uint8_t state = 0;
void setup()

  if (ai.begin(ALGO_OBJECT_DETECTION, MODEL_EXT_INDEX_1)) // Object detection and pre-trained model 1
    Serial.print("Version: ");
    Serial.print("ID: ");
    Serial.println( ai.id());
    Serial.print("Algo: ");
    Serial.println( ai.algo());
    Serial.print("Model: ");
    Serial.print("Confidence: ");
    state = 1;
    Serial.println("Algo begin failed.");

void loop()
  if (state == 1)
    uint32_t tick = millis();
    if (ai.invoke()) // begin invoke
      while (1) // wait for invoking finished
        CMD_STATE_T ret = ai.state(); 
        if (ret == CMD_STATE_IDLE)

     uint8_t len = ai.get_result_len(); // receive how many people detect
       int time1 = millis() - tick; 
       Serial.print("Time consuming: ");
       Serial.print("Number of people: ");
       object_detection_t data;       //get data

       for (int i = 0; i < len; i++)
          Serial.print("Detecting and calculating: ");
          ai.get_result(i, (uint8_t*)&data, sizeof(object_detection_t)); //get result

       Serial.println("No identification");
      Serial.println("Invoke Failed.");
    state == 0;

Open the serial monitor and set baud rate as 115200 and the result of people detection should be showed continuously. Meanwhile, the image that is captured by the module also will display on the website. The successful ongong output should be like:

On the display website, you can see that two faces are framed with dual number. Number 0 means the human face that module detected, and the other number means the confidence.


Not all browsers support the display of Vision AI results. Please refer to the table below for the support of each major browser.

Chrome Edge Firefox Opera Safari
Support Support Not support Support Not support

Customize Trainning Model

The seeed studio has provided pre-trained people detected model and will provide more models in the foreseeable future. You can directly use them to quickly get to know AI camera.

If you want to customize your own firmware, you can refer to here which is powered by YOLOV5.

Or refer to the tutorial here and train your own model to suit your needs.

Burn Training Model into the Module

The training models are all represented as ".uf2" file, which is what we need.

  • Step 1. Connect module to the host PC with a Type-C cable and double click BOOT button on the module.

There will be a "GROVEAI" disk pop up.

  • Step 2. Re-flash the factory firmware

Make sure that the Vision AI's factory firmware has not been modified before flashing the model file. If you have tried other Wiki content, such as capturing images etc., please re-flash the factory firmware before using the model firmware.

Download Grove_AI_default_firmware.


When your model fails to boot or there is a problem with the device, you can try to resolve it by re-flashing Grove_AI_default_firmware.uf2.

  • Step 3. Please copy the uf2 file to the GROVEAI disk to complete the firmware flash.

If you need to flash the factory firmware, copy the factory firmware to the GROVEAI first, then reconnect the Vision AI and finally flash the model firmware.

If you only need to flash the model firmware, just copy the model firmware directly into GROVEAI.

We can see that Work LED on the module flash in speed, that means the process is ongoing. After the disk disapper, the process of burnning firmware is finished.

Update BL702 Chip Firmware

If the Grove Vision AI is not recognised on your computer, please try to recover it by following these steps.

This is the firmware that controls the BL702 chip that builds the connection between the computer and the Himax chip. The latest version of the firmware has now fixed the problem of Vision AI not being able to be recognised by Mac and Linux .

  • Step 2. Download and open BLDevCube.exe software, select "BL702/704/706”, and then click "Finish”
  • Step 3. Click “View”, choose “MCU" first. Move to “Image file”, click “Browse” and select the firmware you just downloaded.
  • Step 4. Make sure there are no other devices connect to the PC. Then hold the Boot button on the module, connect it to the PC.

We can see 5V light and 3.3V LED light are lighting on the back of the module, then loose the Boot button.

  • Step 5. Back to the BLDevCube software on the PC, click "Refresh" and choose a proper port. Then click “Open UART” and ”Creat&Program”, wait for the process done.

Trouble Shooting

  1. What is the main method of connection?

  2. a. Link the microcontroller and the module first.

  3. b. Connect the microcontroller with host PC second.
  4. c. Connect the module with host PC at last.

  5. Why can't I see the image display on the website? Nor the detecting?

  6. a. Is the module heated? Connect the module only will cost heat problem and let the module not functioning for a moment.

  7. b. Is connecting after the correct method?

  8. Why does Grove Vision AI suddenly get an "Algo begin failed." error one day while in use?

  9. In older versions of the initial firmware, it was confirmed that there may be a problem with models being lost during operation. Please follow the steps below and re-flash the firmware to resolve this issue.

  10. Step 1: Connect the Vision AI to the computer and quickly press the BOOT button twice, at which point a disk named GROVEAI will appear on the computer.

  11. Step 2: Please download following file. grove_ai_without_crc_v01-30.uf2
  12. Step 3: Simply copy the file you downloaded in step 2 into GROVEAI.

Tech Support

Please do not hesitate to submit the issue into our forum.