AI, computer vision meet LoRaWAN with SenseCAP K1100 sensor prototype kit

CNXSoft: This is another tutorial using SenseCAP K1100 sensor prototype kit translated from CNX Software Thai. This post shows how computer vision/AI vision can be combined with LoRaWAN using the Arduino-programmable Wio Terminal, a Grove camera module, and LoRa-E5 module connecting to a private LoRaWAN network using open-source tools such as Node-RED and InfluxDB.

Wio Terminal AI Computer Vision

In the first part of SenseCAP K1100 review/tutorial we connected various sensors to the Wio Terminal board and transmitted the data wirelessly through the LoRa-E5 LoRaWAN module after setting the frequency band for Thailand (AS923). In the second part, we’ll connect the Grove Vision AI module part of the SenseCAP K1100 sensor prototype kit to the Wio Terminal in order to train models to capture faces and display the results from the camera on the computer. and evaluate the results of how accurate the Face detection Model is. Finally, we’ll send the data (e.g. confidence) using the LoRa-E5 module to a private LoRaWAN IoT Platform system.

In order to understand what artificial intelligence is, and how it can benefit your business and organization, we should first define a few terms.

  • Artificial Intelligence (AI) is the introduction of human-like intelligence into computers. The goal is to make computers as intelligent as humans by enabling the computer to process specific information (e.g. images, audio), and each time the data is processed, the computer will figure out which outcome is the most likely.
  • Machine Learning (ML) is a sub-category of AI that uses algorithms to automatically learn insights and recognize patterns from data. An algorithm is used together with a model data set (training set) comprised of a number of samples in order to have a ready-to-use result.
  • Deep Learning (DL) goes a step further than machine learning by mimicking the functioning of the human nervous system. It leverages large neural network systems overlapping several layers and learning the sample data. This information can be used to find patterns or categorize the data with the ability to remember faces, customer behavior, etc.

Grove Vision AI module

Grove AI Vision Front
Grove AI Vision (Front)
Grove AI Vision Back
Grove AI Vision (Back)

The Grove AI Vision module comes with a small AI camera supporting TinyML (Tiny Machine Learning) algorithms. It can perform various AI functions provided by Seeed Studio such as person detection, pet detection, people counting, object recognition, etc.. or you could even generate your own model through a training tool for machine learning and take that model. It’s easy to deploy and get results in minutes. The solution does so at ultra-low power with a 2.5mW/frame camera and low-power LoRaWAN connectivity (19.5mW).

Grove AI Vision workflow
Grove Vision AI – Work Flow Diagram
Training Testing machine learning
Train and Test Data set Machine Learning Procedure

 

Grove Vision AI Module Pinout
Grove Vision AI Module specifications

The module also comes with a microphone and a 6-axis motion sensor, so it can be used for more than just AI Vision.

Hardware requirements

We’ll need the following items for our AI vision and LoRaWAN project (items in bold are part of the SenseCAP K1100 kit):

  • Wio Terminal
  • Grove Vision AI Module
  • Grove LoRa-E5 Module
  • 2x USB Type-C cables
  • Computer

Wio Terminal Grove Vision AI LoRaWAN

Deploy the pre-trained AI Vision model and Arduino sketch

Seeed Studio provides pre-trained models that we can use to speed up our learning experience about the Vision AI camera module, including face detection and face & body detection. Let’s see how we can use those.

  • Step 1 – Connect a USB Type-C cable between the Grove Vision AI Module and your computer
    Grove Vision AI Module USB C cable
  • Step 2. Press the BOOT button twice on the Vision AI module to enter Boot mode and it should then appears as the “GROVEAI” drive in your computer with 2 files: INDEX.HTM and INFO_UF2.TXT
    Grove Vision AI Module BOOT buttonGroveAI drive
  • Step 3. We will now select the UF2 firmware (grove_ai_camera_v01-00-0x10000000.uf2) with a pre-trained “People Face Detection” model. Copy the UF2 firmware to the GROVEAI drive, and you should see a blinking light on the Vision AI module indicating it is performing the firmware update.
    Note: If the pre-trained model comes with multiple files, copy one file at a time. Start with the first file, wait until the flashing light goes out, then enter Bootloader mode again and copy the next sequence of files until all are complete.
  • Step 4. While the Vision AI module is still connected to the computer, connect it to the Wio Terminal with a Grove I2C cable, and connect a USB type-C cable to the Wio Terminal controller.

    Grove AI Vision USB Port Windows
    The COM4 is the Wio Terminal, and COM8 is the Grove Vision AI module

  • Step 5. Add the Vision AI module Arduino library by downloading Seeed-Grove-Vision-AI-Moudle.zip. Now go to the Arduino IDE’s top menu and select Sketch -> Include Library -> Add .ZIP Library, and select Seeed-Grove-Vision-AI-Moudle
  • Step 6. Add the LoRa-E5 module Arduino library by downloading Disk91_LoRaE5.zip. Then select Include Library -> Add .ZIP Library, and select Disk91_LoRaE5
  • Step 7. Copy the LoRa-VisionAI-FaceDetection.ino sketch into the Arduino IDE:
  • Step 8. Make sure to edit the program for LoRaWAN configuration:
    • 8 bytes DevEui number
    • 8 bytes AppvEui number
    • 16 bytes APPKEY number
    • Set up an OTAA (Over-The-Air Activation) connection.
    • Set the frequency band for your country For example, I used AS923 in Thailand.

    LoRa Vision AI Face Detection

  • Step 9. Upload the program to the Wio TerminalWio Terminal CNX Software demo
  • Step 10. Open the Serial Monitor to monitor the output from the face detection algorithm, notably two parameters:
    • “Number of people”: the number of people, or technically faces, detected.
    • “Confidence”: the confidence of the face detection algorithm in percentage.

    LoRa Vision AI Face Detection Serial Monitor output

  • Step 11. Visit https://files.seeedstudio.com/grove_ai_vision/index.html with either Google Chrome or Microsoft Edge browser to check the output from the camera
    SenseCAP AI Vision browser
  • Step 12. Click on the Connect button in the SenseCAP AI Vision web page, select Grove AI-Paired, and click Connect.
    Grove AI Paired
  • Step 13. Hold the Grove Vision AI module and point the camera at the people or an image with people. The recognized faces/persons will show up in the web browser in a read box together with a confidence level in percent.
    LoRaWAN AI Vision SenseCAP K1100 Arduino
    Screenshot showing the camera can detect the faces of 3 children with a confidence level of over 74%.

    SenseCAP K1100 Face Detection
    Wio terminal, Vision AI module, LoRa-E5, Arduino serial monitor, and SenseCAP AI Vision website.

Here’s a short demo showing the Grove Vision AI module in action.

Private LoRaWAN IoT On-Premise Platform

So far, we’ve only demonstrated computer vision, but we’ve yet to make use of the LoRa-E5 module. We’ll rely on the same “open-source-powered” private LoRaWAN IoT platform as in the first part of our review, except the environmental sensor is replaced with the Grove Vision AI module.

Private LoRaWAN network AI face detection

  • ChirpStack open-source LoRaWAN network and application server that registers the LoRaWAN IoT device number and decrypts the received data in AES128 format through an MQTT broker acting as the sender (publish).
    ChipStack SenseCAP K1100 Grove AI Vision module
    Received packets

    ChipStack Face Detection data payload
    Payload

  • Node-RED flow-based development tool for programming. It is a recipient (subscribe) from ChirpStack via the MQTT protocol and takes the data from the payload and decodes it according to the BASE64 format. It will store the sensor data in an InfluxDB database.
    Node RED Grove AI Vision module LoRaWAN
  • InfluxDB open-source time series database is used to store the sensor and LoRaWAN gateway data, and automatically sorts it by time series allowing us to analyze the data for any period of time.InfluxDB Studio SensorCAP face detection confidence
  • Grafana real-time dashboard allows the user to visualize the data from the InfluxDB database, this time the number of people and inference confidence.Grafana LoRaWAN People Detection

Final words

  • I’ve now made myself familiar with the SenseCAP K1100 sensor prototype both with the Grove sensors, the Arduino-programmable Wio Terminal, as well as the LoRa-E5 module to send the sensor data to the LoRaWAN Gateway. The more I play with it, the more fun I have, as there are many analog or digital sensors you can use through the 40-pin GPIO header or ready-made sensor modules using I2C or SPI.
  • The Wio Terminal board can be programmed automatically from the Arduino IDE without pressing any buttons. And there are issues entering the Automatic Bootloader mode, Seeed Studio has designed the firmware to also support a Manual Bootloader mode that can be entered by sliding the switch twice quickly. Instructions to enter the manual bootloader mode can be found in the first part of the review.
  • I had to rely on my knowledge and experience with LoRaWAN to transmit data wirelessly over long distances. That’s because LoRaWAN has a more complicated connection process than Wi-Fi or Bluetooth. It would be great if Seeed Studio could develop a new firmware that can connect without any coding
  • The SenseCAP K1100 is a highly capable and reliable kit. It can be especially useful for primary, secondary, and tertiary education and can be integrated into teaching materials in a STEM (Science, Technology, Engineering, and Mathematics) curriculum.

While the SenseCAP K1100 prototype sensor kit can be used to combine AI/computer vision with LoRaWAN, and it is great for education and prototyping, if you plan to deploy such a solution in the field, an industrial-grade sensor should be selected instead. It should be able to withstand the elements such as rain, heat, and dust, and be much more reliable. One example is the industrial-grade SenseCAP A1101 LoRaWAN Vision AI Sensor ($79 on Seeed Studio).

SenseCAP A1101

I would like to thank Seeed Studio for sending the SenseCAP K1100 sensor prototype kit for this review. It is available for $99.00 plus shipping.

Share this:

Support CNX Software! Donate via PayPal or cryptocurrencies, become a Patron on Patreon, or buy review samples

ROCK Pi 4C Plus
Subscribe
Notify of
guest
The comment form collects your name, email and content to allow us keep track of the comments placed on the website. Please read and accept our website Terms and Privacy Policy to post a comment.
1 Comment
oldest
newest
XIAO RP2040 free shipping