Related Experiment Video
Updated: Aug 4, 2025
12:39
A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
Published on: January 18, 2020
7.7K
Non-Intrusive Real Time Eye Tracking Using Facial Alignment for Assistive Technologies
Summary
This study introduces a lightweight, accurate eye-tracking system using convolutional neural networks and a webcam. It offers a faster, more accessible alternative for assistive technologies, improving gaze estimation on mobile devices.
Area of Science:
- Computer Vision
- Human-Computer Interaction
- Assistive Technology
Background:
- Traditional eye tracking systems are often intrusive (head-mounted cameras) or limited by environmental factors (infrared reflections).
- Existing methods pose challenges for long-term use in assistive technologies and can be unreliable in varying light conditions.
Purpose of the Study:
- To develop an accurate and lightweight eye-tracking solution for assistive tasks.
- To enable gaze estimation using a simple webcam and convolutional neural networks.
Main Methods:
- Utilized state-of-the-art convolutional neural network face alignment algorithms.
- Employed a standard webcam for gaze, face position, and pose estimation.
- Focused on appearance-based gaze estimation for improved accessibility.
Main Results:
- Achieved significantly faster computation times (up to 91% decrease) compared to current state-of-the-art methods.
- Maintained comparable accuracy with average errors of 4.5° (MPIIGaze), 3.9° (UTMultiview), and 3.3° (GazeCapture).
- Demonstrated the feasibility of accurate gaze estimation on mobile devices.
Conclusions:
- The proposed webcam-based eye-tracking system is a viable, efficient, and accurate solution for assistive technologies.
- This approach overcomes the limitations of intrusive and environmentally sensitive traditional systems.
- Enables broader application of gaze estimation, including on resource-constrained mobile platforms.
