etdc 0.0.0

Tool for eye-tracking data collection

Eye Tracking Calibration Dataset

• Eye Tracking Calibration Dataset
  • Overview
  • Installation
  • Usage
  • Dataset Contents
  • Methodology
  • Potential Use Cases
  • Acknowledgments

Overview

This dataset is a comprehensive collection of eye-tracking calibration data gathered from multiple participants. It is designed to support research and development in fields such as gaze estimation, eye-tracking systems, human-computer interaction, and computer vision. The dataset includes raw images, facial landmarks, and calibration metadata, making it a versatile resource for training machine learning models and conducting gaze analysis studies.

Installation

pip install etdc

Usage

etdc collect --upload --api-key=<XXXXX>

Dataset Contents

1. Metadata (JSON):
   1. dataset.json:
      1. Landmark coordinates for each frame.
      2. Calibration dot positions and corresponding distances.
      3. Estimated distance of face from camera.
      4. Dot position ( place where eyes are looking at)
   2. screen_details.json:
      1. Screen and window size information to aid in experimental replication.

Methodology

R = 50, 70 cm \\
N = 10

1. Collection Process:
   1. Participants focused on points in a 5x5 grid displayed on a screen.
   2. Eye alignment, eyes position on screen and distance R were continuously monitored during data collection.
   3. N samples were recorded for each calibration point.
2. Participant Diversity:
   1. Data was collected from individuals of various age groups to ensure broad applicability.

Potential Use Cases

1. Training gaze estimation models for real-time applications.
2. Developing assistive technologies for people with mobility impairments.
3. Conducting behavioral studies using gaze patterns.
4. Advancing research in cognitive science, computer vision, and HCI.

Acknowledgments

This dataset was collected using a Python-based tool that leverages:

1. MediaPipe FaceMesh for landmark detection.
2. OpenCV for camera integration and visualization.

We thank all participants who contributed their data to make this dataset a valuable resource for the research community.