FaceGen

FaceGen generates parameterized line drawings of human faces to provide a demographically diverse, quantitative face-space (parametric face drawings, PFDs) for studies of facial recognition and feature analysis.

Key Features:

• Demographic Diversity: Uses a dataset of 400 faces (100 East Asian/Pacific Islander, 100 Latinx/Hispanic, 100 black/African-American, 100 white/Caucasian; 200 female, 200 male) aggregated from databases including FERET and the Chicago Face Database.
• Parameterization and Customization: Encodes each face with 85 landmark points that are normalized and rendered using MATLAB to produce smooth, parameterized line drawings and supports adding faces, morphs, and caricatures.
• Parametric Face Space (PFDs): Implements a face-space model termed parametric face drawings (PFDs) representing statistical distributions of facial features.
• Scientific Validation: Validation comprises two behavioral experiments: Experiment 1 reports an inversion effect in short-term recognition, and Experiment 2 compares celebrity recognition performance of PFDs with untextured FaceGen Modeller renderings (≈50% correct).
• Export Capabilities: Enables export of generated faces for animation, rendering, or 3D printing.

Scientific Applications:

• Identity recognition studies: Supports experiments of short-term and celebrity recognition and assessment of holistic processing via inversion effects.
• Facial-feature analysis and morphing: Facilitates creation of morphs and caricatures for parametric investigations of feature salience and perception.
• Cross-demographic research: Enables comparative studies of recognition and feature perception across East Asian/Pacific Islander, Latinx/Hispanic, black/African-American, and white/Caucasian groups.

Methodology:

Faces are encoded with 85 landmark points, normalized, and modeled as statistical distributions within a parametric face-space (PFDs) and rendered as parameterized line drawings using MATLAB.