SMPL Normal Map Is All You Need for Single-view Textured Human Reconstruction

TL;DR

This paper introduces SEHR, a novel framework for single-view textured human reconstruction that leverages SMPL normal maps and a pretrained 3D model to improve accuracy without relying on diffusion models.

Contribution

The paper proposes the SEHR framework, integrating SMPL normal maps and a pretrained 3D model, to enhance single-view human reconstruction accuracy and efficiency.

Findings

SEHR outperforms existing methods on benchmark datasets.

Incorporates SMPL normal maps for better shape guidance.

Effectively reconstructs invisible body parts with constraints.

Abstract

Single-view textured human reconstruction aims to reconstruct a clothed 3D digital human by inputting a monocular 2D image. Existing approaches include feed-forward methods, limited by scarce 3D human data, and diffusion-based methods, prone to erroneous 2D hallucinations. To address these issues, we propose a novel SMPL normal map Equipped 3D Human Reconstruction (SEHR) framework, integrating a pretrained large 3D reconstruction model with human geometry prior. SEHR performs single-view human reconstruction without using a preset diffusion model in one forward propagation. Concretely, SEHR consists of two key components: SMPL Normal Map Guidance (SNMG) and SMPL Normal Map Constraint (SNMC). SNMG incorporates SMPL normal maps into an auxiliary network to provide improved body shape guidance. SNMC enhances invisible body parts by constraining the model to predict an extra SMPL normal Gaussians. Extensive experiments on two benchmark datasets demonstrate that SEHR outperforms existing state-of-the-art methods.