Architecture

Overview of Webgazer.ts architecture and design decisions.

Core Components

• Trackers - Face and eye feature detection (TensorFlow.js FaceMesh)
• Regressors - Gaze prediction algorithms (Ridge regression variants)
• Filters - Smoothing and noise reduction (Kalman filters)
• Storage - Data persistence (IndexedDB/localStorage)
• Events - Event system for lifecycle management

Design Principles

1. TypeScript-first - Full type safety
2. Modular - Pluggable trackers and regressors
3. Privacy-focused - Opt-in data storage
4. Performance - Web Workers, efficient algorithms
5. Developer experience - Clear APIs, good documentation

More Details

Coming soon! See the source code for implementation details.