MVDream: Multi-view Diffusion for 3D Generation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Shi, Yichun, Wang, Peng, Ye, Jianglong, Long, Mai, Li, Kejie, Yang, Xiao
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909173074100224
author Shi, Yichun
Wang, Peng
Ye, Jianglong
Long, Mai
Li, Kejie
Yang, Xiao
author_facet Shi, Yichun
Wang, Peng
Ye, Jianglong
Long, Mai
Li, Kejie
Yang, Xiao
contents We introduce MVDream, a diffusion model that is able to generate consistent multi-view images from a given text prompt. Learning from both 2D and 3D data, a multi-view diffusion model can achieve the generalizability of 2D diffusion models and the consistency of 3D renderings. We demonstrate that such a multi-view diffusion model is implicitly a generalizable 3D prior agnostic to 3D representations. It can be applied to 3D generation via Score Distillation Sampling, significantly enhancing the consistency and stability of existing 2D-lifting methods. It can also learn new concepts from a few 2D examples, akin to DreamBooth, but for 3D generation.
format Preprint
id arxiv_https___arxiv_org_abs_2308_16512
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle MVDream: Multi-view Diffusion for 3D Generation
Shi, Yichun
Wang, Peng
Ye, Jianglong
Long, Mai
Li, Kejie
Yang, Xiao
Computer Vision and Pattern Recognition
We introduce MVDream, a diffusion model that is able to generate consistent multi-view images from a given text prompt. Learning from both 2D and 3D data, a multi-view diffusion model can achieve the generalizability of 2D diffusion models and the consistency of 3D renderings. We demonstrate that such a multi-view diffusion model is implicitly a generalizable 3D prior agnostic to 3D representations. It can be applied to 3D generation via Score Distillation Sampling, significantly enhancing the consistency and stability of existing 2D-lifting methods. It can also learn new concepts from a few 2D examples, akin to DreamBooth, but for 3D generation.
title MVDream: Multi-view Diffusion for 3D Generation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2308.16512