
Eyes in Motion
A low-cost gaze interface for assistive mobility.
I built a real-time eye-tracking system that uses an ordinary webcam to translate gaze into movement commands, then continued it as my CS senior thesis: letting users draw paths with their gaze instead of issuing constant commands, while the system reads the room around them. The project grew from a personal interest in preserving independence for people whose mobility is limited.
- Earned 3rd Place in the undergraduate ACM Student Research Competition at Tapia 2025 with the first version, tested on a JetBot
- Rebuilt the tracker with MediaPipe and a small custom CNN trained on about 5 minutes of personal calibration data
- Selected the correct cell of a 3x3 grid in 112 of 135 trials
- Added SegFormer floor, wall, and person segmentation with movement-direction cues for the people in view







