Calimiro AI

A powerful smart mirror & mobile app to film and track your workout sessions with AI.

Built in Munich, Germany, by Lennard Gross, Philip Synowiec und Younes Bendimerad at CHECK24

Our goal

Calimiro AI develops a smart mirror (and a mobile app), which is able to track your workout session and to recognize the workout exercise you're doing. It can also count repetitions, and so establish reliable statistics about your global performance. We want to make fitness and calisthenics easier for people who are starting in this discipline, without forcing them to be in the gym or to have a coach.

Project overview

A Calimiro consists of several components. A Raspberry Pi 4+, a transparent mirror with integrated display, and an ESP Camera. A webserver permits a WiFi connection between the Raspberry Pi and the Camera to establish an HTTP (or on newer versions RTSP) live video stream. The core of Calimiro, Calitracker, runs on the Raspberry Pi and analyzes your position with several pre-trained AI models. Another server runs on the Raspberry Pi to permit communication with the User Interface, built with MagicMirror. Since september 2025, we've also been building a mobile app so that the users can run Calimiro on their phone!

Future updates & improvements
• Websockets for a better architecture
• Built-in computation of calories
• Pose & gesture correction
• Workout Session Program generation with AI
• 3D pose & gesture models on the mirror
• Interactive plots of workout data
• Integrate the models into firebase for the mobile app
• Participate to the 'Bundeswettbewerb für KI', the german federal championships for AI

Gallery

App Screenshots

Getting started

The full source code of Calimiro is available on GitHub.