arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2312.11577cs.CV
收藏

PR-NeuS:一种用于快速多视图神经表面重建的基于先验的残差学习范式

PR-NeuS: A Prior-based Residual Learning Paradigm for Fast Multi-view Neural Surface Reconstruction

Huazhong University of Science and Technology(华中科技大学);Nanyang Technological University(南洋理工大学);A*STAR(新加坡科技研究局)

Jianyao Xu, Qingshan Xu, Xinyao Liao, Wanjuan Su, Chen Zhang, Yew-Soon Ong, Wenbing Tao

AI总结:

针对多视图神经表面重建训练耗时长的问题,提出PR-NeuS基于先验的残差学习范式,通过两阶段优化结合先验引导采样,实现3分钟单场景重建且质量具竞争力。

摘要:

神经表面学习在多视图表面重建任务中已展现出优异性能。但现有大多数方法采用大型多层感知器(MLPs)从零开始训练模型,单个场景的训练耗时长达数小时。近年来,如何加速神经表面学习受到广泛关注,目前仍是一个待解决的开放问题。\n本文提出一种基于先验的残差学习范式,用于快速多视图神经表面重建。该范式包含两个优化阶段:第一阶段,我们提出利用泛化模型生成基础有符号距离函数(SDF)场,通过融合泛化模型生成的多个局部SDF场即可快速得到该初始场,为重建提供粗糙的全局几何先验。第二阶段基于该先验,提出一种基于哈希编码网络的快速残差学习策略,为基础SDF场编码一个偏移SDF场。此外,我们引入先验引导的采样方案,帮助残差学习阶段实现更好的收敛,从而恢复更精细的结构。\n实验结果表明,采用本文设计的范式,我们的方法仅需约3分钟即可完成单个场景的表面重建,同时能达到具有竞争力的表面质量。我们的代码将在论文发表后公开。

英文摘要:

Neural surfaces learning has shown impressive performance in multi-view surface reconstruction. However, most existing methods use large multilayer perceptrons (MLPs) to train their models from scratch, resulting in hours of training for a single scene. Recently, how to accelerate the neural surfaces learning has received a lot of attention and remains an open problem. In this work, we propose a prior-based residual learning paradigm for fast multi-view neural surface reconstruction. This paradigm consists of two optimization stages. In the first stage, we propose to leverage generalization models to generate a basis signed distance function (SDF) field. This initial field can be quickly obtained by fusing multiple local SDF fields produced by generalization models. This provides a coarse global geometry prior. Based on this prior, in the second stage, a fast residual learning strategy based on hash-encoding networks is proposed to encode an offset SDF field for the basis SDF field. Moreover, we introduce a prior-guided sampling scheme to help the residual learning stage converge better, and thus recover finer structures. With our designed paradigm, experimental results show that our method only takes about 3 minutes to reconstruct the surface of a single scene, while achieving competitive surface quality. Our code will be released upon publication.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑