NAMASTE
An immersive yoga experience
Inspired by the literal names of yoga poses, I created Namaste to help users understand the origins of pose names and become one with their environment.
THE PROJECT
WHAT IS NAMASTE?
A live yoga-teaching program meant to serve as an introduction to basic yoga and expose people to this form of medication and exercise.
Namaste was my final project for the class, Machine Learning and New Interfaces. Instructions were to use the programs/tools we learned in class (ie: Bodypix, PoseNet, KNNClassifier, etc.)
WHAT DOES IT DO?
On a computer, user's start off in a virtual background with a given list of yoga pose names and corresponding demonstrations. They must guess which pose is related to that background and perform the pose accordingly (ie: a dog park would relate to "downwards dog").
Striking an 100% accurate pose will allow users to move on to a different background, and the process repeats.
HOW DOES IT WORK?
Based on a database I created, the computer recognizes each yoga pose a user performs through machine learning. The accuracy of the pose is determined by how similar a user's pose is to the correlating pose in the database. An 100% accurate pose triggers a 3 second timer that is used to transition between the different environments.
Users appear in virtual backgrounds through a body-segmentation program (Bodypix), which can differentiate people from the surrounding real-life environment. Whatever Bodypix recognizes as not body will not be shown on the screen from a live webcam feed.
GALLERY
This project was presented at the Fall 2019 IMA end-of-semester show, which is open to the school and public.
TAKEAWAYS
Machine learning and yoga aside, this project was an unexpected lesson about human behavior.
I was surprised to observe that many adults appeared hesitant/embarrassed to perform yoga in public, whereas kids were less self-conscious to do so. However, once the adults saw other adults doing yoga, they eagerly joined.
Moving forward, I must not only work on the technical functions of a project but also take into account the audience's comfort level and environment the work will be presented in.
TOOLS USED
ML5.js
BODYPIX
ATOM
ADOBE PHOTOSHOP
This machine learning library was used to create a database of different yoga forms that was then used to asses the accuracy of the poses user's struck in front of the laptop camera.
Bodypix was used for body-segmentation to create a green-screen effect, where only the user's body would appear in the virtual environment, rather than the entire physical environment.
This text editor program was used to code HTML, CSS, and Javascript.
The images of the yoga poses were photoshopped to remove the original background.