Gaze Tracking at 100 Hz
Novel approach to real-time gaze estimation using temporal fusion and geometric constraints. Achieves professional-grade performance without specialized hardware or GPU acceleration. View demonstration.
Novel approach to real-time gaze estimation using temporal fusion and geometric constraints. Achieves professional-grade performance without specialized hardware or GPU acceleration. View demonstration.
Bridging the gap between laboratory precision and deployable systems through algorithmic innovation
Full 3D gaze direction recovery with off-camera angle estimation. Enables precise attention modeling and eye-contact quantification beyond binary classification schemes.
Lightweight computational pipeline achieving 100 Hz sampling on commodity hardware. Eliminates GPU dependency through optimized geometric algorithms and efficient feature extraction.
Complete on-device processing with no data transmission. Architectural approach ensures GDPR compliance and enables deployment in privacy-sensitive contexts.
Multi-frame temporal integration compensates for noise and illumination variability. Maintains stable tracking under challenging real-world conditions including low light and rapid movement.
Functions with conventional RGB webcams without infrared illumination or depth sensors. Rapid calibration procedure enables immediate deployment across diverse user populations.
Simultaneous extraction of gaze coordinates, confidence estimates, head pose parameters, and ocular events. Provides rich behavioral signals for downstream analysis.
Combining visual, acoustic, and interaction signals for comprehensive behavioral assessment
Real-time audio stream analysis for ambient activity detection. Enables contextual awareness in remote assessment and collaborative environments through acoustic pattern recognition.
Temporal analysis of interaction patterns including typing velocity, pause distributions, and input rhythm characteristics. Provides engagement proxies for attention state inference.
On-device speech recognition for accessibility applications and real-time transcription. Supports note-taking workflows and verbal interaction documentation in educational contexts.
Probabilistic fusion of gaze vectors, acoustic events, and input patterns to estimate engagement levels. Provides holistic behavioral signatures for activity classification.
All audio and keystroke processing occurs locally with configurable retention policies. Raw sensor data never leaves the device, ensuring compliance with data protection regulations.
Performance characteristics from controlled evaluation protocols
Sustained high-frequency tracking with 10ms temporal resolution
Complete independence from cloud infrastructure and network connectivity
Rapid deployment from SDK initialization to production streaming
Validated deployments across telecommunication, education, and automotive sectors
Multi-modal behavioral monitoring for remote assessment environments. Integrates visual attention, acoustic context, and interaction patterns for comprehensive engagement analysis.
Driver monitoring without infrared illumination requirements. Enables eyes-off-road detection through efficient CPU-based processing.
Exploring research partnerships and technology transfer opportunities