| Host | link |
|---|---|
| Github | 🔗 |
| NBviewer | 🔗 |
The goal of this project was to estimate where the eye gaze of the user was positionned in space
Here is the result:
This is a python notebook, so you can easily launch it on your computer.
It's pretty straightforward.
- Use Dlib facial landmark detection for face position and eye extraction
- Use Fabian Timm methods to find pupils.
- Find eye rotation matrix (using pupils and eye corners) and face rotation matrix (from facial landmarks) using PNP solving (see here)
- Draw the gaze vector on screen
Of course for a precise application you need to calibrate your camera parameter and surface position, but the gist of it is here!