Inspiration
The inspiration comes rather random.
What it does
It has every function possible of a mouth that you can trigger through doing specific facial gesture.
How I built it
I used a model that could detect faces from a picture and used another one that could detect 68 landmarks of the face (on those faces that i just extracted). The rest is pure creativity and trials and errors.
Challenges I ran into
It is a lot harder to simulate the mouse with your face since your face can hardly stay very still. It will alwasy got a shakiness. Not to mention the different in lighting and angle of the face that might cause it to be misdetected and thus, misfunctioned.
Accomplishments that I'm proud of
Still i managed to got most of the basic with the accuracy of the task (given that the face is always frontal) up to 90%.
What I learned
What's next for Facial Mouse Controller
I think if i got the resources for training some of my own models for detecting part of the face and then used each of them in the specific task, it would increase the efficient of this by many times.
Log in or sign up for Devpost to join the conversation.