A biologically plausible model of human shape symmetry perception

Area of Science:

• Computational neuroscience
• Visual perception
• Computer vision

Background:

• Symmetry perception is crucial for object recognition but computationally challenging.
• Existing models often struggle to reliably encode symmetry.
• Previous work established models for shape perception and symmetry data.

Purpose of the Study:

• To extend an existing shape perception model to incorporate shape symmetry encoding.
• To investigate the neural mechanisms underlying symmetry perception.
• To achieve human-comparable performance in symmetry detection.

Main Methods:

• The study extends a computational model of shape perception.
• The model incorporates object-centric symmetry mechanisms.
• It utilizes V4-like concentric units for position recovery and multiplicative curvature mechanisms for shape encoding.

Main Results:

• The extended model achieves performance comparable to humans in perceiving shape symmetry.
• The model requires minimal additional neural circuitry beyond existing shape perception components.
• The model shows potential for encoding symmetry axes in natural images with improved edge recovery.

Conclusions:

• A unified computational model can account for both shape and symmetry perception.
• The proposed model offers an efficient approach to encoding visual symmetry.
• The findings suggest a neural basis for symmetry processing in the visual system.