Hi3DEval: Advancing 3D Generation Evaluation with Hierarchical Validity
Yuhan Zhang, Long Zhuo, Ziyang Chu, Tong Wu, Zhibing Li, Liang Pan, Dahua Lin, Ziwei Liu
2025-08-08
Summary
This paper talks about Hi3DEval, a new method to judge how good 3D models and objects created by AI are. It checks both the whole object and its individual parts to see if they look real, including the materials they are made of.
What's the problem?
The problem is that existing ways to evaluate AI-generated 3D content don't look closely enough at the details and different parts of objects, which means they can miss problems with the quality or realism of the models.
What's the solution?
The solution was to design Hi3DEval, a system that uses a large dataset and combines different types of 3D data. It evaluates models step by step from overall shape to small details and materials, giving a more thorough and accurate assessment.
Why it matters?
This matters because better evaluation helps improve 3D generation technology, making AI-created models more realistic and useful for things like games, movies, and virtual reality.
Abstract
Hi3DEval is a hierarchical evaluation framework for 3D generative content that combines object-level and part-level assessments, including material realism, using a large-scale dataset and hybrid 3D representations.