How Do Users Map Points Between Dissimilar Shapes?

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Computational Geometry

Background:

• Extensive research exists on finding corresponding points in similar shapes using descriptors and matching methods.
• However, identifying similar points in dissimilar shapes remains an underexplored area.

Purpose of the Study:

• To investigate how humans identify similar points between dissimilar 2D shapes.
• To develop predictive models for user point mappings based on geometric relationships.
• To enable content transfer between disparate shapes.

Main Methods:

• User study involving mapping points between dissimilar 2D shapes.
• Analysis of user mappings to identify correlations with geometric properties.
• Development of two statistical models to predict user mapping probability distributions.
• Validation and comparison of predictive models against established shape-matching techniques.

Main Results:

• User mappings between dissimilar shapes strongly correlate with simple geometric relationships.
• The developed statistical models accurately predict the probability distribution of user mappings.
• Proposed models outperform or match existing shape-matching methods in predicting human perception.

Conclusions:

• Geometric relationships are key to human point correspondence in dissimilar shapes.
• Statistical models based on geometry can effectively predict user perception.
• This approach facilitates object and procedural content mapping across diverse shapes for design applications.