Reliable Data Collection Methodology for Face Recognition in Preschool Children

Area of Science:

• Biometrics
• Computer Vision
• Developmental Psychology

Background:

• Existing facial datasets predominantly feature adults, limiting research on children.
• Collecting biometric data from minors requires stringent consent protocols, creating data scarcity.
• Privacy concerns and logistical challenges hinder the development of comprehensive children's face datasets.

Purpose of the Study:

• To address the scarcity of child facial data by creating a novel dataset.
• To investigate optimal camera configurations for accurate face recognition in children.
• To provide guidelines for the ethical and practical construction of children's face datasets.

Main Methods:

• Collected facial data from 74 children aged 2-7 years in daycare settings.
• Conducted experiments with cameras installed in diverse locations to assess recognition performance.
• Analyzed data to identify key considerations for dataset creation and optimal camera placement.

Main Results:

• Successfully established a unique facial dataset of young children.
• Determined specific camera locations and setups that yield superior face recognition accuracy for children.
• Identified critical factors and methodologies for building robust children's face recognition databases.

Conclusions:

• The developed dataset and experimental findings offer valuable resources for child-focused biometrics research.
• Optimal camera placement strategies can significantly enhance the performance of face recognition systems for children.
• This study provides a foundational framework for future research in children's facial recognition technology.