graviti
产品服务
解决方案
知识库
公开数据集
关于我们
avatar
ICG Graz50
2D Box
Urban
|...
许可协议: Unknown

Overview

This is a dataset of rectified facade images and semantic labels.The goal of the annotation is to study the layout of the facades.It contains 50 images of various architectural styles such asClassicism, Biedermeier, Historicism (neo-renaissance andneo-baroque), Art Nouveau and as well as various modern styles.It portrays a wider range of complex facade layouts than other datasets.The images are generated automatically after extracting a piecewise planar geometryfrom about 30 perspective images each. The view is orthographic however thepiecewise planar geometry is only an approximation, which leads toartifacts during the image blending.The facade images are labeled in multiple classes:labels_used - wall, door, window, sky - are used in the publication (see below)labels_full - shop, balcony, roof - are NOT used due to too little training datawhere the wall class is very coarse and covers mostly all backgroundwhich is not window or door, including sky and void areas of the image.At the time of publication, the performance was the following:Window Wall Door Sky | Global Class IoUMAP 60 66 57 80 66 65 43Our 60 84 41 91 78 69 58The protocol is to use 30 images for training and 20 for testing.The dataset is used for facade layout study as the facade is mostlysegmented out clean. Hence the goal is not to label all classes as best as possibleyet to understand the types of layout of different architectural styles.If you use the datasets, please cite:Irregular lattices for complex shape grammar facade parsing. CVPR 2012, Providence, USA.H. Riemenschneider, U. Krispel, W. Thaller, M. Donoser, S. Havemann, D. Fellner, H. Bischof.http://www.vision.ee.ethz.ch/~rhayko/paper/cvpr2012_complex_facades_hayko.pdf

数据概要
数据格式
image,
数据量
--
文件大小
--
| 数据量 -- | 大小 --
ICG Graz50
2D Box
Urban
许可协议: Unknown

Overview

This is a dataset of rectified facade images and semantic labels.The goal of the annotation is to study the layout of the facades.It contains 50 images of various architectural styles such asClassicism, Biedermeier, Historicism (neo-renaissance andneo-baroque), Art Nouveau and as well as various modern styles.It portrays a wider range of complex facade layouts than other datasets.The images are generated automatically after extracting a piecewise planar geometryfrom about 30 perspective images each. The view is orthographic however thepiecewise planar geometry is only an approximation, which leads toartifacts during the image blending.The facade images are labeled in multiple classes:labels_used - wall, door, window, sky - are used in the publication (see below)labels_full - shop, balcony, roof - are NOT used due to too little training datawhere the wall class is very coarse and covers mostly all backgroundwhich is not window or door, including sky and void areas of the image.At the time of publication, the performance was the following:Window Wall Door Sky | Global Class IoUMAP 60 66 57 80 66 65 43Our 60 84 41 91 78 69 58The protocol is to use 30 images for training and 20 for testing.The dataset is used for facade layout study as the facade is mostlysegmented out clean. Hence the goal is not to label all classes as best as possibleyet to understand the types of layout of different architectural styles.If you use the datasets, please cite:Irregular lattices for complex shape grammar facade parsing. CVPR 2012, Providence, USA.H. Riemenschneider, U. Krispel, W. Thaller, M. Donoser, S. Havemann, D. Fellner, H. Bischof.http://www.vision.ee.ethz.ch/~rhayko/paper/cvpr2012_complex_facades_hayko.pdf

0
立即开始构建AI
graviti
wechat-QR
长按保存识别二维码,关注Graviti公众号

Copyright@Graviti
沪ICP备19019574号
沪公网安备 31011002004865号