WisBlock is a modular system that makes it easy to implement edge IoT solutions. It is your ally from rapid prototyping to mass production without the need to create new hardware modules for each step.
WisBlock ecosystem lets you decide on several core options as the RAK11200 (ESP32), RAK4631 (nRF52840), and the RAK11310 (RP2040), to develop your TinyML project alongside the largest variety of sensors in the market, from PDM microphones for audio recognition to motion sensors for movement models and many more.
A starter kit is available in the RAK store.
Setting Up Your Development Board
The minimum hardware you will need to start developing with WisBlock and Edge Impulse are as follows:
- WisBlock Base
- WisBlock RAK11200 (ESP32) or RAK11310 (RP2040) core
- RAK1904 3-Axis Acceleration Sensor (LIS3DH) or RAK18000 PDM Microphone
Installing Dependencies
To set WisBlock up in Edge Impulse, you will need to install the following software:
Uploading Data to Edge Impulse
Once you have the software already installed, it’s time to start creating your dataset.
Select the core you are using:
Deploying the Trained Model Locally
Once your dataset is created and your model is trained, it’s time to deploy it back to your WisBlock.
Select the core you are using:
RAK11200 Edge Impulse Guide
RAK11200 Data Uploading
Download the Edge Impulse firmware .zip file.
Download the ESP Flash Download Tool.
Open the Flash Download Tool and set chipType to "ESP32" and left workMode on "develop", then click OK.
Connect the WisBlock Base, where RAK11200 is attached to the computer via USB.
Short the BOOT0 and GND pins, then reset the board to restart the core in boot mode.
Search for the firmware files and set them on the right memory partition.
Search for the ei_rak11200_firmware.bin and set it on @ 0x10000 partition
Search for the bootloader.bin and set it on @ 0x1000 partition
Search for the partition-table.bin and set it on @ 0x8000 partition
Set the SPI SPEED to 40 MHz, SPI MODE to DIO, and DoNotChgBin checked. Everything should look like this:
Click the START button and wait until you see a FINISH status on the Download panel.
Finally, reset the board.
| 📝 NOTE: You can also build the firmware using the RAK11200 Edge Impulse source code. |
Connecting to Edge Impulse
Connect your WisBlock board to your PC through the USB cable. In the command prompt or terminal, run the following code:
edge-impulse-daemon
This will start a wizard which will ask you to log in and choose an Edge Impulse project. If you want to switch projects, run the command:
--clean
Verifying Connection
After running the command, your WisBlock is now connected to Edge Impulse. To verify this, go to your Edge Impulse Project, and click Devices. The device should be listed here.
Building a Machine Learning Model
Now that you have successfully set up the connection, you can now build your first machine learning model. For this, you need to connect supported WisBlock sensors to your board. In this case, a PDM microphone or a 3-axis accelerometer.
Collecting your first data:
Define the label and the sample length of your data, and click Start sampling. After the time you have set, you will see the uploaded sample.
Now, you are ready to create your dataset for your custom project. Also, you can follow these tutorials for specific use cases.
| 📝 NOTE: The Data forwarder lets you easily send data from any sensor into Edge Impulse. For more WisBlock Sensors options, refer to RAK store. |
WisBlock team is working on supporting new sensors with Edge Impulse everyday.
RAK11200 Trained Model Deployment
RAK11200 ESP32 core lets you deploy your model using the native ESP IDF or PlatformIO. You can select your preferred option.
Deploying Your Model Back to WisBlock
ESD IDF
After training and validating your model, you can now deploy it to your WisBlock board. This makes the model run without an internet connection, minimizes latency, and runs with minimal power consumption.
On your Edge Impulse project, navigate to the Deployment tab on the left menu and select the C++ library option.
Edge Impulse offers an optimization tool that can help you increase on-device performance but may reduce accuracy. You can test each option one by one and stay with the best results. You can also click the Analyze optimizations button before deploying it and see the recommended choice for your target. Or, click Build to use the currently selected option.
Download the .zip library and save it to your preferred location, then unzip it.
Now, you need to build the firmware for your corresponding core with the trained model added to it.
Download and install ESP IDF v4.4.
- Download or clone the firmware from the Github repository.
- You'll need two additional components to compile the firmware:
- LIS3DHTR_ESP-IDF
- ESP-NN
- Then open the ESP-IDF 4.4 CMD from the Desktop shortcut.
- Navigate to the firmware folder
../RAK11200-Edge-Impulse-Firmware.
Get the additional components by cloning the corresponding repositories to the components folder in the root folder of the project: ../RAK11200-Edge-Impulse-Firmware/components.
cd components
git clone <https://github.com/AIWintermuteAI/LIS3DHTR_ESP-IDF.git> LIS3DHTR_ESP-IDF/
cd LIS3DHTR_ESP-IDF
git checkout 641bda8c3e4b706a2365fe87dd4d925f96ea3f8c
cd ..
git clone <https://github.com/espressif/esp-nn.git> esp-nn/
cd esp-nn
git checkout 24d18025f300c1e15afa2abb86519da54c7a5d90
cd ..
cd ..
With the last cd .., you must be in the project folder ../RAK11200-Edge-Impulse-Firmware.
Add Your Trained Model Library
Replace the edge-impulse-sdk, model-parameters, and tflite-model folders from the project with the new ones in the trained model library.
Build the Application
- Delete the build folder from the project. (Important)
- Run the following command:
idf.py build
Flash
- Connect the RAK11200 ESP32 WisBlock board to your computer.
- Short BOOT0 and GND then press the Reset button of the base board. (To restart the core in boot mode)
- Identify the COM port where it's connected.
- Run the following command:
idf.py -p COMxx flash monitor
Where COMxx needs to be changed to the actual port where your WisBlock is connected on your system.
- After uploading successfully, reset the board.
| 📝 NOTE: Also, you can build and flash the firmware using the Visual Studio Code Esspressif plugin. |
Run the Model
From a command prompt or terminal, run:
edge-impulse-run-impulse
If your model uses audio, you may use the --continuous command as follows:
edge-impulse-run-impulse --continuous
Your model should run and print the inference results in the terminal continuously.
PlatformIO
After training and validating your model, you can now deploy it to your WisBlock board. This makes the model run without an internet connection, minimizes latency, and runs with minimal power consumption.
| 📝 NOTE: In installing and setting up the PlatformIO to work with WisBlock ecosystem boards, refer to the Learn section. |
On your Edge Impulse project, navigate to the Deployment tab on the left menu and select the Arduino library option.
Edge Impulse offers an optimization tool that can help you increase on-device performance but may reduce accuracy. You can test each option one by one and stay with the best results. You can also Analyze optimizations before deploying it and see the recommended choice for your target. Or, click Build to use the currently selected option.
Download the .zip library, save it to your preferred location, and then unzip it.
Finally, to deploy your model:
- Clone or download the WisBlock examples for PlatformIO from the repository.
- Open the example that better fits your needs on Visual Studio Code.
If you are working with:
- Keyword spotting or sound recognition → “inference_microphone_continuous” is recommended.
- Accelerometer or any other sensor → “inference_accelerometer” is recommended.
Drag and drop the Arduino library to the lib folder of your project:
Finally, Build your project, connect your WisBlock board to your computer and Upload the code.
Run the Serial Monitor, and you should start seeing the inference results:
RAK11310 Edge Impulse Guide
RAK11310 Data Uploading
Download the Edge Impulse firmware .uf2 file.
Connect the WisBlock Base where RAK11310 is attached to the computer via USB.
Hold the BootSel button in the RAK11300 module, reset the module (press the reset button of WisBlock Base Board for at least 2 seconds), then wait for the USB Mass Storage RPI-RP2 to appear before releasing the BootSel button.
Drag and drop the .uf2 file to the RPI-RP2 USB Mass Storage.
| 📝 NOTE: You can also build the firmware using the RAK11310 Edge Impulse source code. |
Connecting to Edge Impulse
Connect your WisBlock board to your PC through the USB cable. From a command prompt or terminal, run the following command:
edge-impulse-daemon
This will start a wizard which will ask you to log in and choose an Edge Impulse project. If you want to switch projects, run the command:
--clean
Verifying Connection
After running the command, your WisBlock is now connected to Edge Impulse. To verify this, go to your Edge Impulse Project, and click Devices. The device should be listed here.
Building a Machine Learning Model
Now that you have successfully set up the connection, you can now build your first machine learning model. For this, you need to connect supported WisBlock sensors to your board. In this case, a PDM microphone or a 3-axis accelerometer.
Collecting your first data:
Define the label and the sample length of your data, and click Start sampling. After the time you have set, you will see the uploaded sample.
Now, you are ready to create your dataset for your custom project. Also, you can follow these tutorials for specific use cases.
| 📝 NOTE: The Data forwarder lets you easily send data from any sensor into Edge Impulse. For more WisBlock Sensors options, refer to RAK store. |
WisBlock team is working on supporting new sensors with Edge Impulse everyday.
RAK11310 Trained Model Deployment
Deploying Your Model Back to WisBlock
After training and validating your model, you can now deploy it to your WisBlock board. This makes the model run without an internet connection, minimizes latency, and runs with minimal power consumption.
On your Edge Impulse project, navigate to the Deployment tab on the left menu and select the C++ library option.
Edge Impulse offers an optimization tool that can help you increase on-device performance but may reduce accuracy. You can test each option one by one and stay with the best results. You can also Analyze optimizations before deploying it and see the recommended choice for your target. Or, click Build to use the currently selected option.
Download the .zip library and save it to your preferred location, then unzip it.
Now, you need to build the firmware for your corresponding core with the trained model added to it.
- For Windows users, follow this guide: https://shawnhymel.com/2096/.
- For macOS users, follow this guide: https://datasheets.raspberrypi.com/pico/getting-started-with-pico.pdf
| 📝 NOTE: Tested with the GNU Arm Embedded Toolchain version 10.3.1. |
- Download or clone the firmware from the Github repository.
Add Your Trained Model Library
Replace the edge-impulse-sdk, model-parameters, and tflite-model folders from the project with the new ones in the trained model library.
Build the Application
- Delete the build folder from the project. (Important)
- Run the following command:
mkdir build
cd build
cmake -G "MinGW Makefiles" ..
make
Flash the Firmware
Connect the RAK11310 RP2040 WisBlock board to your computer.
Hold the BootSel button in the RAK11300 module, reset the module (press the reset button of WisBlock Base Board for at least 2 seconds), then wait for the USB Mass Storage RPI-RP2 to appear before releasing the BootSel button.
Run the following command:
cp ei_rak11310_firmware.uf2 /g/
After uploading successfully, reset the board.
| 📝 NOTE: Also, you can build and flash the firmware using the Visual Studio Code CMake tools. |
Run the Model
From the command prompt or terminal, run:
edge-impulse-run-impulse
If your model uses audio, you may use the --continuous command as follows:
edge-impulse-run-impulse --continuous
Your model should run and print the inference results in the terminal continuously.
RAK4631 Edge Impulse Guide
The RAK4631 does not have firmware like the other cores to upload data with the official Edge Impulse ingestion service, but the solution for this is covered in this guide.
RAK4631 Data Uploading
Audio Data
To upload audio directly to Edge Impulse, you can use your smartphone or personal computer. This way, you can create your dataset even easier. On your Edge Impulse project, go to Devices and click Connect a new device:
Select the device you want to use, mobile phone or computer:
Once you select your device, the Audio data collection web interface will appear, then grant access to your microphone:
Then, give a name to your sample, and define its length and category, then start recording:
You should see a success message pops up, and your recording should be visible in the Data acquisition tab of your project:
You can easily record high-quality and long-lasting audio recordings using this method. Enjoy creating your dataset.
Other Sensor Data
If your project uses a custom or any other sensor from our ecosystem, use the data forwarder to upload your data through serial communication.
If, for example, you are working with a 3-axis accelerometer with a simple code, make your WisBlock print the axis data separated by commas at a certain frequency.
Serial.print(x);
Serial.print(",");
Serial.print(y);
Serial.print(",");
Serial.println(z);
The output should look like this:
-0.12,-6.20,7.90
-0.13,-6.19,7.91
-0.14,-6.20,7.92
-0.13,-6.20,7.90
-0.14,-6.20,7.91
Once your board is printing the data through serial, open a command prompt or terminal and run the data forwarder by typing:
edge-impulse-data-forwarder
If this is the first time running this command, you will be asked to enter your Edge Impulse account credentials. Select the project you want to send your data to and name the axis. The baud rate should be 115200, but if you want to use a different one, you need to change it like this:
edge-impulse-data-forwarder --baud-rate 460800
For examples and more options, refer to this guide: Data Forwarder.
RAK4631 Trained Model Deployment
After training and validating your model, you can now deploy it to your WisBlock board. This makes the model run without an internet connection, minimizes latency, and runs with minimal power consumption.
On your Edge Impulse project, navigate to the Deployment tab on the left menu and select the Arduino library option.
Edge Impulse offers an optimization tool that can help you increase on-device performance but may reduce accuracy. You can test each option one by one and stay with the best results. You can also Analyze optimizations before deploying it and see the recommended choice for your target. Or, click Build to use the currently selected option.
Download the .zip library, save it to your preferred location, and then unzip it.
| 📝 NOTE: In installing and setting up the PlatformIO to work with WisBlock ecosystem boards, refer to the Learn section. |
Finally, to deploy your model, execute the following steps:
- Clone or download the WisBlock examples for PlatformIO from the repository.
- Open the example that better fits your needs on Visual Studio Code.
If you are working with:
- Keyword spotting or sound recognition → “inference_microphone_continuous” is recommended.
- Accelerometer or any other sensor → “inference_accelerometer” is recommended.
Finally, Build your project, connect your WisBlock board to your computer and Upload the code.
Run the serial monitor, and you should start seeing the inference results:
| 📝 NOTE: RAK4631 works only with PlatformIO. |
Updated