GazeProjector: Accurate Gaze Estimation and Seamless Gaze Interaction Across ...
summary
GazeProjector maps 2D pupil positions from a Pupil Labs head-mounted eye tracker into display coordinates by matching streamed display screenshots to the scene camera with natural feature tracking and a homography, so gaze projects onto any screen. A prototype runs on a central server and is tested with 12 participants across a projection wall, Surface tabletop and iPad, outperforming marker-tracking and head-orientation baselines with significantly lower gaze error and enabling no re-calibration across displays, but it requires streamed screenshots, feature-rich content and stable lighting.
AI-written summary · about
chapters (9)
- 0:00Introduction to GazeProjector
- 1:51Problem with current eye trackers
- 3:37Related work overview
- 5:44System implementation details
- 8:13Single display evaluation
- 11:04Multi-display evaluation
- 13:32Limitations of the system
- 14:23Summary and contributions
- 15:52Q&A session
related talks
19mEveryday Eye Contact Detection Using Unsupervised Gaze Target Discovery
19mEyeScout: Active Eye Tracking for Position and Movement Independent Gaze Interaction with Large ...
20mBuilding a Personalized, Auto-Calibrating Eye Tracker from User Interactions
19mAggreGaze: Collective Estimation of Audience Attention on Public Displays
17mEyeGrip: Detecting Targets in a Series of Uni-directional Moving Objects Using Optokinetic ...
22mPinpointing: Precise Head- and Eye-Based Target Selection for Augmented Reality
09mUsing Gaze Behavior and Head Orientation for Implicit Identification in Virtual Reality
16mEye-Trace: Segmentation of Volumetric Microscopy Images with Eyegaze