EyeTeractive: Paraconsistent Logic for Gaze-Based Interfaces

Overview

The EyeTeractive project aims to develop an accessible and efficient gaze-tracking system by leveraging paraconsistent logic and artificial intelligence. This system translates eye movements into computational commands, providing inclusive solutions for users with physical or motor limitations. By utilizing cost-effective hardware like standard webcams and robust algorithms, the project overcomes the challenges of traditional gaze-tracking technologies.

Key Features

• Paraconsistent Logic: Handles uncertainties and inconsistencies in gaze-tracking data caused by variable lighting and image quality.
• AI-Powered Processing: Uses the ResNet architecture and Dlib library for accurate iris detection and gaze direction estimation.
• Accessibility: Eliminates the need for expensive hardware like Tobii Eye Trackers, enabling compatibility with low-cost devices.
• Real-Time Interaction: Processes and maps gaze movements into actionable commands in real time.

Motivation

Traditional gaze-tracking solutions often require proprietary hardware that is expensive and not universally accessible. The EyeTeractive project addresses these issues by:

1. Reducing dependency on specialized hardware.
2. Enhancing robustness in challenging environments (e.g., variable lighting).
3. Promoting digital inclusion by making the technology accessible to diverse user groups.

Objectives

General Objective

• Develop an efficient model to translate gaze positions into computational commands using paraconsistent logic and artificial intelligence.

Specific Objectives

1. Validate the model's performance under adverse conditions, such as lighting and viewing angles.
2. Ensure scalability and applicability in low-cost devices.

Methodology

1. Data Collection:

   • 7,000+ frames were collected from 27 videos of volunteers at the Federal University of Rondônia.
   • Participants followed a moving circle on a screen to capture gaze data.

2. Model Development:

   • Utilized the Dlib library for facial landmarks and iris region detection.
   • Trained the model using the ResNet architecture for robust classification and detection tasks.
   • Avoided data augmentation to ensure real-world scenario validity.

3. Paraconsistent Logic Integration:

   • Applied paraconsistent logic to process and translate gaze data, addressing inconsistencies effectively.
   • Calculated precise gaze directions using horizontal and vertical displacement metrics.

Visualization

• Calculating the Horizontal Axis:

  

  This image illustrates the calculations used to determine the horizontal displacement of the iris.

• Eye Position Interpretation:

  

  This graph visualizes the processed interpretation of the eye's position.

• Adapted Paraconsistent Logic:

  

  This chart represents the application of adapted paraconsistent logic to interpret gaze directions.

• Iris Mask Extraction:

  

  This image shows the segmented mask of the eye region extracted using the Dlib library.

Applications

• Biometric analysis.
• Eye-based interface control for individuals with motor impairments.
• Behavioral studies using gaze tracking.

Dependencies

• Python 3.8+
• Dlib
• OpenCV
• TensorFlow or PyTorch (for ResNet implementation)

Getting Started

1. Clone the repository:

   git clone https://github.com/user/EyeTeractive.git
   

2. Install Dependencies:

   pip install -r requirements.txt
   

3. Run the application:

   python main.py
   

Team

• Jáder Louis de Souza Gonçalves
• Samih Santos de Oliveira

Advisor

• Prof. Dr. Lucas Marques da Cunha