Augmented mirror

Description

This project uses multiples recognition softwares to project the body pose and features of the user on a mirror matching his viewpoint.

Requirements

• Nvidia-Docker
• A nvidia GPU with drivers installed
• Chromium web browser

Using the Docker container

To build the container from the docker file, simply use make build.
To execute the container, use make run. Make sure that you've already installed nvidia-docker and that your computer supports cuda.
Inside the docker container launch_reflection.sh will be executed and launch the recognition software.

(Optional) Manualy sending position data to the server

In /reflections, while being in the docker container, use:

python3 send_data.py

Openning the app

You have to use a simple local http server to host the index.html and all the files. For exemple, use python3 -m http.server in the apps/ folder.

Use make open to see the result.

Ressources:

• https://github.com/openpifpaf/openpifpaf
• https://google.github.io/mediapipe/solutions/holistic.html
• https://google.github.io/mediapipe/solutions/hands.html
• https://github.com/IntelRealSense/librealsense/
• https://github.com/socketio/socket.io
• https://p5js.org/