Assessing 3D shape cognition using BVH-based hierarchical geometric disclosure

Abstract

Assessing human 3D shape cognition in a controlled and quantifiable manner remains challenging, as existing approaches often rely on visual cues influenced by prior experience. We introduce a geometry-based experimental paradigm in which 3D objects are progressively disclosed using a Bounding Volume Hierarchy (BVH), a widely used hierarchical structure in computer graphics. By revealing bounding volumes level by level, the framework provides a controllable and reproducible mechanism for regulating the amount of geometric information available during each trial. A user study across multiple object categories shows that recognition responses frequently emerge at coarse-to-intermediate BVH levels and approach saturation at deeper levels, while recognition accuracy increases as additional geometric detail becomes available. Higher confidence is associated with correct recognition outcomes and, less robustly, with responses at shallower BVH levels. These results suggest that hierarchical geometric structure provides an effective basis for controlled evaluation of human 3D shape recognition. BVHs thus offer a structured framework for constructing controlled geometric stimuli and for probing recognition thresholds, accumulation dynamics, and the relationship between confidence and accuracy in human 3D shape cognition.