#IMPRESSIVE
Everyday toolIoTAIIntegrationInnovation
Handy

Pythonscikit-learnMediaPipeFlutterHome Assistant
Language: English + PolishEasily control your home automations using hand gestures, at any time. This app recognizes the gestures using your security camera footage and controls every IoT action you can dream of. Nothing is impossible - make your life feel like you were the main character of a sci-fi movie.
The video says everything! (sound needed)
Features

🖐️ Gesture Recognition

Control your smart home devices effortlessly with various hand gestures - I built and trained a classifier (Random Forest) model by myself. It gets the angles between different body joints (thanks to MediaPipe and some magic (math to be specific)) as input data and outputs the corresponding gesture ID. This sophisticated model accurately interprets your hand movements, enabling seamless interaction with your IoT devices.

🏡 Home Assistant Integration

Handy was built to seamlessly integrate with Home Assistant, the most popular central hub for your smart home ecosystem. This means you can effortlessly control all your IoT devices that are compatible with it. Every single one of them. There is no catch.

🎵 Built-In Actions

The core element of Handy is music control - play, pause, adjust volume, switch tracks, play music from YouTube or online radio, and much more. Want to get an update on weather or want the music player to say the current date and time? No problem at all.

📱 Mobile App

If you need to manually control Handy without using hand gestures, you can use the mobile app I also created. What's more, it keeps the action history so that you can see who did what and when. You can also manage music items (radios/YouTube videos) of the playlist used in Handy.

⚙️ Resource Efficiency

Efficiently operates on standard surveillance cameras 24/7, without any problems. Optimized for fully automated workflow - you forget that this app "works somewhere" - these hand gestures become a crucial part of your life (trust me, I know what I'm talking about 😅)

🤖 Customization

The app can be customized in every way. You can even build, train, and test your own gesture recognition model - I wanted this project to be incredibly friendly for new users - see below for Jupyter notebooks on how to get started. Easily customize and extend actions for any IoT device - with the intuitive code and docs.

📈 Constant Innovation

Since this is a crucial part of my (and my family members') lives right now, every problem that I discover will be patched. And of course, this doesn't stop just there. This app will be improved so that it becomes a real science fiction dream come true.

How was this project created?Recently, I've been really interesting in machine learning and computer vision. The primary reason behind my projects in general is to address some problem, often related to human shortcomings, like laziness. The same was true for Handy - outside my house, we have a music player (Volumio) that usually plays online radio. However, sometimes it's too loud, sometimes too quiet, and sometimes it would be helpful to know the time or check the weather, and so on. Nobody in my family wants to rush inside the house to grab their phone (for example when working in the garden) just to check these things. That's why I came up with the idea of creating something that, using existing hardware (the total cost of this project is literally $0, with no need to install anything new), namely a surveillance camera and the music player I mentioned, would make this process easier. Handy recognizes gestures that allow you to control devices in my home IoT network, including that player. Was there a problem? Yes. Is there a solution? Always.
Made with ❤️ by Artur Nowakjust like every project on this page ✨