Quality Assessment of 3D Digital Humans Constructed from Captured Data： Datasets and Methods

Digital humans have recently emerged as a prominent form of digital media and have increasingly contributed to the advancement of digital technology in China. Digital humans primarily fall into two categories： those made using measured data and those produced with generative tools. Although generative approaches offer rapid and flexible design pipelines， their practical adoption remains limited due to persistent technical constraints. In contrast， 3D digital humans created with data captured from real people using high-precision sensing devices and expert postprocessing continue to dominate industrial applications. Nevertheless， the performance of digital humans made with captured data is constrained by the available sensing and communication infrastructure. Thus， they tend to exhibit various forms of quality degradation. Consequently， quality assessment has become essential for improving the production and user experience of 3D digital humans made with captured data. Thus， methods to assess the quality of these systems have attracted considerable attention in recent years as a topic of active research. This paper provides the first systematic review of this emerging field. We examine design and production pipelines used to create 3D digital humans with captured data and analyze potential quality distortions that can be introduced at each stage. We then summarize representative datasets that are commonly used as subjective and objective quality assessment methods. Finally， we discuss current challenges and highlight possible directions for future research.