Fast and Unassisted Eye Tracker Calibration for Gaze-Based Pervasive Human-Computer Interaction
summary
CalibMe is an unsupervised, marker-driven calibration that uses detected fiducial markers (ArUco-style) and automatic outlier removal on pupil size/position to collect many calibration points from a moving cellphone display. Evaluated on head-mounted trackers, it achieves 0.59° mean angular error (σ=0.23) versus 0.82° (σ=0.15) for nine-point calibration, with typical calibration times around 10–20 s for novices and experts; implemented on a practical system and shown to drive gaze-based actions like microscope autofocus.
AI-written summary · about
chapters (11)
- 0:00Introduction to eye tracking
- 1:09Current calibration challenges
- 3:18Traditional calibration process
- 4:23Proposed calibration method
- 5:46Marker design and detection
- 7:09Outlier removal techniques
- 8:56Automatic evaluation metrics
- 10:45Experimental setup and results
- 13:56Conclusion and open-source tool
- 14:38Q&A: User distraction during calibration
- 15:22Q&A: Remote eye tracker adaptation
related talks
19mEyeScout: Active Eye Tracking for Position and Movement Independent Gaze Interaction with Large ...
06mCaptivates: A Smart Eyeglass Platform for Across-Context Physiological Measurement
18mMatchPoint: Spontaneous Spatial Coupling of Body Movement for Touchless Pointing
09mRotoWrist: Continuous Infrared Wrist Angle Tracking using a Wristband
18mFinexus: Tracking Precise Motions of Multiple Fingertips Using Magnetic Sensing
07mAuraRing: Precise Electromagnetic Finger Tracking
12mEclipseTouch: Touch Segmentation on Ad Hoc Surfaces using Worn Infrared Shadow Casting
05mReady, Steady, Touch! — Sensing Physical Contact with a Finger-Mounted IMU