Edge Impulse Object Detection Demo (Arduino App Lab)

This project demonstrates real-time object detection using Edge Impulse models and a Flask web interface. Designed for Arduino App Lab and used for the MWC Barcelona 2026 workshops.

Features

• Run Edge Impulse object detection models on images, videos, RTSP streams, or connected cameras
• Live web UI for viewing results and switching sources
• Easy setup for both Arduino App Lab and local development

Quick Start (Arduino App Lab)

1. Open the Arduino App Lab and Connect your Arduino UNO Q.
2. SSH connexion: Click on the "terminal" icon on the bottom-left corner:

3. Go to the ArduinoApp/ directory:

cd home/arduino/ArduinoApps/

4. Clone this repo into your ArduinoApps folder:

   git clone https://github.com/edgeimpulse/example-arduino-app-lab-object-detection-using-flask.git
   cd example-arduino-app-lab-object-detection-using-flask/

5. In Arduino App Lab, go to My Apps (top-left corner). You should see the new application:

6. Make the models executable (or add your own):

   chmod +x models/rubber-ducky-linux-aarch64.eim 
   chmod +x models/rubber-ducky-fomo-linux-aarch64.eim

7. Start the app:

   arduino-app-cli app start .

8. View logs (optional):

   arduino-app-cli app logs .

9. Stop the app:

   arduino-app-cli app stop .

The logs will provide something similar to:

...
Edge Impulse runner initialized.
Server running at: http://172.18.0.2:5001
* Serving Flask app 'main'
* Debug mode: on

Note that this IP address is the internal docker network address. To get the local IP address of your UNO Q, use the following command:

hostname -I
192.168.1.8 172.17.0.1 172.18.0.1 2a01:e0a:c6:14b0:c428:1d78:47e0:1f8e 

10. Open a web browser using your local IP address, here in this case 192.168.1.8:5001

Local Development

1. Install Python 3.13.5 (recommended: pyenv):

   pyenv install 3.13.5
   pyenv local 3.13.5

2. Set up a virtual environment:

   python -m venv .venv
   source .venv/bin/activate

3. Install dependencies:

   pip install --upgrade pip
   pip install -r python/requirements.txt

4. Run the app:

   python python/main.py

Usage

• Open your browser to the address shown in the terminal (default: http://localhost:5001)
• Use the web UI to select input source (image, video, RTSP, or camera)
• View original and detection results side by side

Notes

• Place your images/videos in the assets/ folder.
• Place your Edge Impulse .eim models in the models/ folder.
• Only macOS (arm64) and Linux (aarch64) are supported for model inference.