Measuring the Visual Salience of 3D Printed Objects.

Area of Science:

• Computer Vision and Human-Computer Interaction
• Visual Perception and Cognitive Science

Background:

• Understanding human visual attention is crucial for fields like computer graphics, virtual reality, and robotics.
• Previous research on visual saliency has primarily focused on 2D stimuli, leaving a gap in understanding attention towards 3D objects.
• Validating computational models of visual saliency requires robust experimental data on real-world object viewing.

Purpose of the Study:

• To investigate human viewing behavior and fixation patterns on physical 3D objects.
• To validate existing assumptions about visual saliency in the context of 3D stimuli.
• To develop methods for comparing fixation data across subjects and for generating control fixation sequences.

Main Methods:

• Utilized eye-tracking technology combined with a scene camera and fiducial markers on 3D objects.
• Collected fixation data from human observers viewing physical 3D stimuli.
• Developed algorithms for comparing fixation sequences and generating random fixation data for control purposes.

Main Results:

• Demonstrated significant agreement in human observers' fixation points on the same 3D object under similar viewing conditions.
• Established a procedure for validating computational models of visual saliency for 3D objects.
• Found that prevalent mesh saliency models, based on center-surround patterns, do not accurately predict human fixations on 3D objects.

Conclusions:

• Human visual attention towards 3D objects exhibits predictable patterns and inter-observer agreement.
• Current computational models for 3D visual saliency are inadequate and require further development.
• The developed methodology provides a framework for future research into 3D visual attention and model validation.