WebEyeTrack

Calibrated Real-Time Eye Tracking

RUN

This tool integrates RedForestAI/WebEyeTrack into the EZ-MMLA Toolkit as a browser-based webcam eye-tracking workflow.

WebEyeTrack combines MediaPipe face landmarks with a TensorFlow.js gaze model to estimate a point of gaze on the screen. This Toolkit version adds a guided two-round calibration routine: one round on a white screen and one round on a black screen, with 25 target points in each round.

The run page follows the same Toolkit workflow as the other webcam tools: camera capture, tagging, recording controls, downloadable CSV predictions, and an overview panel with session statistics and a gaze scatter plot.

Output Description
x, y Estimated gaze position in screen pixels.
norm_x, norm_y WebEyeTrack normalized point of gaze in the -0.5 to 0.5 range.
gaze_state Whether the model considers the eyes open or closed.
face_visible Whether face landmarks were detected for that sample.
total_ms Total per-frame processing time reported by WebEyeTrack.

Source Code