Credits
What this site is built with
The site relies on open libraries and models written by others. Here is the list, with the licence of each one and the link to its source. All files are served by this site: the browser downloads nothing from other addresses.
Recognition of the hand, body and face
- MediaPipe Tasks Vision 1.0.1 by Google (Apache 2.0 licence): the library that, inside the browser, finds the landmarks of the hand, body and face in every frame.
- The MediaPipe models
hand_landmarker,gesture_recognizer,pose_landmarker_liteandface_landmarkerby Google (Apache 2.0 licence): the pre-trained neural networks the library uses. The model cards are in the MediaPipe documentation.
Graphics and page building
- three.js 0.186 (MIT licence): the 3D hand and skeleton.
- Vite (MIT licence) and TypeScript (Apache 2.0 licence): they build the code of the pages.
- Node.js (MIT licence): the small server that delivers the files.
The school course
- p5.js (LGPL 2.1 licence) and its web editor: the environment where students write their code.
- ml5.js 1.4.0 (MIT licence), with Google's HandPose model: the hand recognition used in the lessons.
- The three versions of the course site are built with MkDocs 1.6.1 (BSD 2 licence) and the Material for MkDocs 9.7.7 theme (MIT licence), with Quarto 1.10.19 (MIT licence, with Bootstrap and Bootstrap Icons, MIT licence) and with mdBook 0.5.4 (MPL 2.0 licence, with the Open Sans and Source Code Pro fonts, highlight.js and Font Awesome, each with its own free licence).
- The emoji in the MkDocs version of the course are Twemoji (graphics under CC BY 4.0 licence).
Sources
- The numbering of the 21 hand landmarks and the 33 body landmarks follows the MediaPipe documentation: hands, body.
- The rules on the webcam in the browser come from MDN: getUserMedia.
Licences of this site
The texts are published under the Creative Commons Attribution 4.0 licence: you may copy and modify them as long as you credit the author. The code is published under the MIT licence.