利用单视图航拍影像学习生成建筑屋顶的三维表示
计算机视觉与模式识别
2023-03-21 v1 机器学习
摘要
我们提出了一种新颖的流程,用于在假设屋顶几何遵循一组规则图案的条件下,学习给定航拍图像像素时建筑屋顶网格的条件分布。不同于需要同一物体多张图像的替代方法,我们的方法仅使用单张图像即可估计三维屋顶网格。该方法采用 PolyGen(一种用于三维网格的深度生成式 transformer 架构)。我们将该模型应用于一个新领域,并研究了图像分辨率的敏感性。我们提出了一种新度量来评估推断网格的性能,结果表明该模型即使在较低分辨率下也具有鲁棒性,同时在定性上为分布外样本生成了逼真的表示。
引用
@article{arxiv.2303.11215,
title = {Learning to Generate 3D Representations of Building Roofs Using Single-View Aerial Imagery},
author = {Maxim Khomiakov and Alejandro Valverde Mahou and Alba Reinders Sánchez and Jes Frellsen and Michael Riis Andersen},
journal= {arXiv preprint arXiv:2303.11215},
year = {2023}
}
备注
Copyright 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works