Scholarlink

<p>Recently, efficiently learning a user-specific neural 3D body model for Novel view synthesis has drawn significant attention. However, existing neural human rendering methods struggle with a single-image input when large disparity captures are encountered. In this paper, we propose IDiff-NeRF, a novel framework for high-fidelity and plausible free-view human body synthesis exclusively from a single image. Our method builds a diffusible density volumetric representation to necessitate identity-based rendering drawing upon the diffusion model. Employing a parametric score to depict the 3D representation distribution, we strive to achieve Pareto optimality in the non-equilibrium optimization framework, balancing realistic rendering and feasible predictions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

IDiff-NeRF: single-view 3D human body reconstruction utilizing identity-based diffusion within implicit neural network framework

  • 期刊论文

  • 发表日期：

    2025年6月13日

查看全文
  • Yansen Huang,
  • Hongji Yang,
  • Jiao Liu,
  • Bo Ren

摘要

Recently, efficiently learning a user-specific neural 3D body model for Novel view synthesis has drawn significant attention. However, existing neural human rendering methods struggle with a single-image input when large disparity captures are encountered. In this paper, we propose IDiff-NeRF, a novel framework for high-fidelity and plausible free-view human body synthesis exclusively from a single image. Our method builds a diffusible density volumetric representation to necessitate identity-based rendering drawing upon the diffusion model. Employing a parametric score to depict the 3D representation distribution, we strive to achieve Pareto optimality in the non-equilibrium optimization framework, balancing realistic rendering and feasible predictions.