Title: A Survey of Recent Progress in Neural 3D Mesh Texturing

Abstract:

The visual fidelity of digital assets and environments is heavily dependent on texturing 3D meshes. While contemporary generative techniques leveraging Neural Radiance Fields and Gaussian Splatting enable the direct creation of textured content, polygonal meshes remain the fundamental format across industries such as gaming, visual effects, animation, and 3D modeling. Consequently, neural approaches to 3D mesh texturing persist as a critical and dynamic field of study. This survey offers an in-depth examination of the latest developments in this domain, focusing on techniques for texture synthesis, transfer, and completion. We begin by outlining the foundational principles of neural generative models, differentiable rendering, texture mapping, and mesh geometry. Subsequently, we categorize existing research into a cohesive taxonomy that ranges from early Generative Adversarial Network (GAN) methodologies to contemporary diffusion-based frameworks. Additionally, the paper evaluates prevalent architectural designs and supervision mechanisms, surveys available datasets and assessment standards, and explores both practical commercial implementations and emerging use cases. By addressing these open challenges and current applications, this work aims to provide a structured overview of the field’s landscape, thereby informing and guiding future research in learning-driven 3D mesh texturing.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC