Edge Impulse Makes TinyML Accessible for Millions of Arduino Developers

Posted by: ArduinoTEAM — May 26th, 2020

Running Machine Learning (ML) on microcontrollers is one of the most exciting developments in recent years, allowing small battery-powered devices to detect complex movements, recognize sounds, or find anomalies in sensor data. To make it accessible for every embedded developer to build and deploy these models, we have launched top-notch support for Edge Nano on the Arduino Nano 33 BLE Sense and other 32-bit Arduino boards.

The trend of running ML on microcontrollers is known as Embedded ML or TinyML. This means that devices can make intelligent decisions without sending data to the cloud—this is great from both efficiency and privacy perspectives. Now, even powerful deep learning models (based on artificial neural networks) can be applied to microcontrollers. Over the past year, significant progress has been made to make deep learning models smaller and faster to run on embedded hardware through projects like TensorFlow Lite Micro, uTensor, and Arm’s CMSIS-NN; however, creating high-quality datasets, extracting the right features, training, and deploying these models is still complex.

With Edge Impulse, you can now quickly collect real-world sensor data, train ML models on the cloud, and then deploy the models back to Arduino devices. From there, you can integrate the model into your Arduino sketch with a single function call. This way, your sensors become smarter and can understand complex events in the real world. Built-in examples allow you to collect data from accelerometers and microphones, but it’s easy to integrate other sensors with just a few lines of code.

Excited? Here’s how to build your first deep learning model using Arduino Nano 33 BLE Sense (there’s also a video tutorial: Setting Up Arduino Nano 33 BLE Sense with Edge Impulse: https://docs.edgeimpulse.com/docs/arduino-nano-33-ble-sense).

Sign up for a free Edge Impulse account!

• Install Node.js and Arduino CLI.

• Insert the Arduino Nano 33 BLE Sense development board.

• Download the Arduino Nano 33 BLE Sense firmware—this is a special firmware package (source code) that contains all the code for quickly collecting data from its sensors. Launch the Flash script suitable for your platform to refresh the firmware.

• Start the Edge Impulse daemon and connect your board to Edge Impulse. Open a terminal or command prompt and run:

$ npm install edge-impulse-cli -g

$ edge-impulse-daemon

• Collect some data and build a model. We have organized two end-to-end tutorials: Detecting gestures using accelerometers or detecting audio events using microphones.

• Now, your device will appear in the “Devices” tab of the Edge Impulse studio, allowing you to collect some data and build a model.

• Once completed, you can deploy the model back to Arduino Nano 33 BLE Sense. It can be a binary file containing the complete ML model or an Arduino library integrated into any sketch.

Deploying from Edge Impulse to Arduino

• Your machine learning model is now running on the Arduino board. Open the serial monitor and run “AT + RUNIMPULSE” to start classifying real-world data!

Discovering Keywords on Arduino Nano 33 BLE Sense

Integrate with Your Favorite Arduino Platform

We have launched the Arduino Nano 33 BLE Sense, but you can also integrate Edge Impulse with your favorite Arduino platform. You can easily collect data from any sensor and development board using a data forwarder. This is a small application that reads data serially and sends it to Edge Impulse. All you need is a few lines of code in your sketch.

(Example: https://docs.edgeimpulse.com/docs/cli-data-forwarder#example-arduino)

Once the model is built, you can easily export it as an Arduino library. This library can run on any Arm-based Arduino platform (including the Arduino MKR series or Arduino Nano 33 IoT) as long as it has sufficient RAM to run your model. Now, you can include the ML model in any Arduino sketch with just a few lines of code. After adding the library to the Arduino IDE, you can find examples for integrating the model under File > Examples > Your Project – EdgeImpulse > static_buffer.

To run your model as quickly and energy-efficiently as possible, we automatically leverage the hardware capabilities of Arduino boards, such as signal processing extensions on the Arm Cortex-M4 based Arduino Nano BLE Sense or the more powerful Arm Cortex-M7 based Arduino Portenta H7. We also utilize the optimized neural network kernels provided by Arm in CMSIS-NN.

Towards Mass Production

We believe that many embedded applications today can benefit from ML, whether for predictive maintenance (this machine starts to behave abnormally), helping workers stay safe (fall detection), or healthcare (detecting early signs of potential infections). By combining Edge Impulse and the Arduino MKR series, you can quickly deploy ML-based simple applications that integrate LoRa, NB-IoT cellular, or WiFi connectivity. Next, we will add integration for Arduino Portenta H7 on Edge Impulse, enabling higher-performance industrial applications.

END

Yinghe Academy

The Yinghe team is dedicated to providing electronic engineers and students in related fields with standardized core skills courses, helping everyone effectively enhance their professional capabilities at various stages of learning and work.

Yinghe Academy

Exploring the electronic field together

Follow the Yinghe public account for instant access to classes

Click to read the original text for more