Augmented Mirror
An interactive augmented reality mirror using computer vision for pose estimation and gesture-based interactions.
Project
An interactive augmented reality mirror using computer vision for pose estimation and gesture-based interactions.
View code ↗01
Context
Second Self explores the mirror as a modular augmented-reality interface. The system detects the person standing in front of it, aligns digital content with their reflection and provides a shared foundation for applications in learning, health and physical training.
02
Pose estimation and alignment
A pose-estimation model detects body, face and hand landmarks. Depth data from an Intel RealSense D435i camera converts image coordinates into real-world coordinates so that graphics can be positioned consistently on the reflected body.
03
System architecture
The computer-vision pipeline runs inside a Docker container for portability. Detected coordinates are transmitted to a server and then to the browser-based display, separating perception, data exchange and rendering.
04
Applications
Prototypes built on the platform include sign-language training, a music-practice module and pose-controlled minigames such as Poke It, Bounce and Dance. Together they test the mirror as an extensible medium rather than a single-purpose product.
