UniLat3D: Geometry-Appearance Unified Latents for Single-Stage 3D Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Guanjun, Fang, Jiemin, Yang, Chen, Li, Sikuang, Yi, Taoran, Lu, Jia, Zhou, Zanwei, Cen, Jiazhong, Xie, Lingxi, Zhang, Xiaopeng, Wei, Wei, Liu, Wenyu, Wang, Xinggang, Tian, Qi
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915522473361408
author Wu, Guanjun
Fang, Jiemin
Yang, Chen
Li, Sikuang
Yi, Taoran
Lu, Jia
Zhou, Zanwei
Cen, Jiazhong
Xie, Lingxi
Zhang, Xiaopeng
Wei, Wei
Liu, Wenyu
Wang, Xinggang
Tian, Qi
author_facet Wu, Guanjun
Fang, Jiemin
Yang, Chen
Li, Sikuang
Yi, Taoran
Lu, Jia
Zhou, Zanwei
Cen, Jiazhong
Xie, Lingxi
Zhang, Xiaopeng
Wei, Wei
Liu, Wenyu
Wang, Xinggang
Tian, Qi
contents High-fidelity 3D asset generation is crucial for various industries. While recent 3D pretrained models show strong capability in producing realistic content, most are built upon diffusion models and follow a two-stage pipeline that first generates geometry and then synthesizes appearance. Such a decoupled design tends to produce geometry-texture misalignment and non-negligible cost. In this paper, we propose UniLat3D, a unified framework that encodes geometry and appearance in a single latent space, enabling direct single-stage generation. Our key contribution is a geometry-appearance Unified VAE, which compresses high-resolution sparse features into a compact latent representation -- UniLat. UniLat integrates structural and visual information into a dense low-resolution latent, which can be efficiently decoded into diverse 3D formats, e.g., 3D Gaussians and meshes. Based on this unified representation, we train a single flow-matching model to map Gaussian noise directly into UniLat, eliminating redundant stages. Trained solely on public datasets, UniLat3D produces high-quality 3D assets in seconds from a single image, achieving superior appearance fidelity and geometric quality. More demos \& code are available at https://unilat3d.github.io/
format Preprint
id arxiv_https___arxiv_org_abs_2509_25079
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle UniLat3D: Geometry-Appearance Unified Latents for Single-Stage 3D Generation
Wu, Guanjun
Fang, Jiemin
Yang, Chen
Li, Sikuang
Yi, Taoran
Lu, Jia
Zhou, Zanwei
Cen, Jiazhong
Xie, Lingxi
Zhang, Xiaopeng
Wei, Wei
Liu, Wenyu
Wang, Xinggang
Tian, Qi
Computer Vision and Pattern Recognition
Artificial Intelligence
Graphics
High-fidelity 3D asset generation is crucial for various industries. While recent 3D pretrained models show strong capability in producing realistic content, most are built upon diffusion models and follow a two-stage pipeline that first generates geometry and then synthesizes appearance. Such a decoupled design tends to produce geometry-texture misalignment and non-negligible cost. In this paper, we propose UniLat3D, a unified framework that encodes geometry and appearance in a single latent space, enabling direct single-stage generation. Our key contribution is a geometry-appearance Unified VAE, which compresses high-resolution sparse features into a compact latent representation -- UniLat. UniLat integrates structural and visual information into a dense low-resolution latent, which can be efficiently decoded into diverse 3D formats, e.g., 3D Gaussians and meshes. Based on this unified representation, we train a single flow-matching model to map Gaussian noise directly into UniLat, eliminating redundant stages. Trained solely on public datasets, UniLat3D produces high-quality 3D assets in seconds from a single image, achieving superior appearance fidelity and geometric quality. More demos \& code are available at https://unilat3d.github.io/
title UniLat3D: Geometry-Appearance Unified Latents for Single-Stage 3D Generation
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Graphics
url https://arxiv.org/abs/2509.25079