IntraFace

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Biometrics

Background:

• Increasing interest in automated facial image analysis for diverse applications.
• Existing commercial software lacks comprehensive, integrated, and freely available solutions.
• Need for advanced tools for facial feature tracking, pose estimation, attribute recognition, and expression analysis.

Purpose of the Study:

• Introduce IntraFace (IF), a publicly available software package for automated facial image analysis.
• Present a novel technique for unsupervised synchrony detection to analyze correlated facial behaviors.
• Provide a comprehensive and accessible tool for researchers and developers.

Main Methods:

• Development of the IntraFace (IF) software package.
• Implementation of algorithms for facial feature tracking, head pose estimation, facial attribute recognition, and facial expression analysis.
• Introduction of a new unsupervised synchrony detection technique for analyzing correlated facial behaviors.

Main Results:

• IntraFace (IF) achieved state-of-the-art results in emotion expression and action unit detection on FERA, CK+, and RU-FACS databases.
• Demonstrated the capability to measure audience reaction to a live talk.
• Successfully discovered synchrony in smiling behaviors within parent-infant interaction videos.

Conclusions:

• IntraFace (IF) offers a powerful, free, and integrated solution for automated facial image analysis.
• The novel synchrony detection method opens new avenues for studying interpersonal facial behavior dynamics.
• IF provides state-of-the-art performance and is valuable for research in computer vision and related fields.