Related Experiment Video
Updated: Sep 16, 2025
12:39
A Methodology for Capturing Joint Visual Attention Using Mobile Eye-Trackers
Published on: January 18, 2020
7.8K
Dual Focus-3D: A Hybrid Deep Learning Approach for Robust 3D Gaze Estimation
Abderrahmen Bendimered1, Rabah Iguernaissi1, Mohamad Motasem Nawaf1
1Laboratoire d'Informatique et des Systèmes, CNRS UMR 7020, Aix-Marseille University, 13009 Marseille, France.
Sensors (Basel, Switzerland)
|July 12, 2025
Summary
This study introduces Dual Focus-3D, a novel deep learning model for accurate gaze estimation by fusing eye images and 3D head pose. It achieves state-of-the-art results on the new EyeLis dataset.
Area of Science:
- Computer Vision
- Machine Learning
- Human-Computer Interaction
Background:
- Gaze estimation is crucial for understanding human attention.
- Existing methods struggle in natural environments.
- Applications range from assistive technology to virtual reality.
Purpose of the Study:
- To develop a robust and accurate gaze estimation model.
- To introduce a novel hybrid deep learning architecture.
- To present a new dataset for training and evaluation.
Main Methods:
- Developed Dual Focus-3D, a hybrid deep learning architecture.
- Fused appearance-based eye features with 3D head orientation data.
- Introduced the EyeLis dataset with 5206 annotated samples.
Main Results:
- Achieved state-of-the-art performance with a Mean Absolute Error (MAE) of 1.64° on the EyeLis dataset.
- Demonstrated effective generalization across synthetic and real datasets.
- Showcased significant accuracy improvements by incorporating 3D spatial information.
Conclusions:
- The Dual Focus-3D model offers enhanced accuracy and robustness in gaze estimation.
- Multimodal feature fusion and 3D spatial information are key to improved performance.
- The EyeLis dataset facilitates further research in 3D gaze prediction.
