Gesture Control

Scope

Code Project · Computer Vision

Client

College Project

Year

[2023]

Industry

Software Development

Apple Vision Pro intrigued me for a while and to be honest, It was enough to get me into this rather difficult (for me) computer vision project. I liked the gesture based controls it had & I liked it even more when it was demonstrated at WWDC.



At the end I did get bored very quickly after trying to get it to move my cursor faster by increasing the framerate at which the model processes images of my hand from my laptop’s webcam. Best I could manage was 60hz tracking with 99% GPU usage. 



Safe to say I’ll not be trying that again..

Boxes of Brand Packaging

Context

This project was inspired by Apple Vision Pro's gesture inputs & other github projects that explored gesture control.

Approach

Computer vision libraries in python were used to train a model that placed & tracked markers on the user's right hand. Gestures were trained for left click, right click & drag.

Outcome

Image throughput from the webcam to the script was heavily GPU bound. Limitations resulted in a 30hz tracking rate & unstable GPU temps.

Credits

Developed by Saraj Raja

Edit Content