Contrastive self-attention for pupil segmentation
A neural-network approach for finding the pupil accurately in eye images used by gaze-tracking systems.
View paper ↗Research across computer vision, healthcare ML, human-computer interaction, and AI systems—connected to the engineering problems it was designed to solve.
A neural-network approach for finding the pupil accurately in eye images used by gaze-tracking systems.
View paper ↗Deep-learning methods for detecting small light reflections in eye images and matching them to the correct sources.
View paper ↗A complete VR eye-tracking pipeline spanning cameras, illumination, image processing, geometry, calibration, and gaze estimation.
View thesis ↗Image-processing methods for isolating retinal exudates that may support diabetic-retinopathy screening.
View paper ↗A device and signal-processing approach for measuring oxygen saturation without per-device calibration.
View paper ↗Early medical-device research on estimating hemoglobin without drawing a conventional blood sample.
View paper ↗An evaluation of how different signal-processing choices affect a calibration-free pulse-oximeter pipeline.
View paper ↗Fuse face information with the most visually useful eye so gaze estimation can run on ordinary RGB-camera devices.
US12175014B2 ↗Estimate 3D gaze with a single visible reflection, pupil position, and head pose instead of requiring several reflections.
WO2024059927A1 ↗Learn person-specific eye geometry during calibration, then simplify the hardware required during everyday tracking.
WO2025113309A1 ↗Use fast eye movement to identify a screen region and a physical pointer to make the final precise selection.
US20240211034A1 ↗Model relationships between wearable signals and blood characteristics relevant to cardiorenal metabolic conditions.
WO2025106664A1 ↗Analyze a phone photo or scan to decide whether an at-home biological sample is likely suitable for laboratory testing.
WO2023091455A1 ↗Multimodal interaction for identifying and selecting meaningful regions during AI-assisted image editing.
Filing not publicMethods for carrying visual concepts across media when configuring a virtual-keyboard theme.
Filing not publicA combined gaze-and-pen interaction method for moving quickly and then positioning a cursor precisely.
Filing not publicA multi-camera approach to using corneal reflections for robust gaze estimation.
Filing not publicA visual explanation of six public families—from mobile gaze tracking and gaze-assisted interaction to wearable biomarkers and at-home sample quality.