Global Urban Network SVI Data

This data set contains segmentation masks for city-scale street view imagery. We employ Meta's Mask2Former approach (Cheng, 2022) for image segmentation. <br> We obtain SVI from Mapillary which is the world's largest platform for free and open street view imagery (SVI). Till date, Mapillary's coverage has penetrated most global cities around the world. Compared to other proprietary options such as Google Street View, Mapillary images are hosted on a CC-BY-SA 4.0 licence, which allows users to freely share, use, and adapt images. <br> Images are subsampled proportionately based on geographic tiling of each city's administrative boundary and filtered to remove non-frontal facing and non-daylight images. <br> <br>

本数据集包含城市级街景影像的分割掩码（segmentation masks）。我们采用Meta公司的Mask2Former方法（Cheng, 2022）进行图像分割工作。
我们从Mapillary获取街景影像（Street View Imagery，简称SVI），该平台是全球规模最大的免费开源街景影像平台。截至目前，Mapillary的覆盖范围已遍及全球绝大多数城市。与谷歌街景（Google Street View）等专有服务相比，Mapillary影像采用CC-BY-SA 4.0许可证发布，允许用户自由分享、使用及修改影像。

影像按照各城市行政边界的地理分块进行比例抽样，并过滤掉非正脸朝向及非日间拍摄的影像。

搜集汇总

背景与挑战

背景概述

该数据集提供全球多个城市的街景图像分割掩码，采用Meta的Mask2Former方法处理，数据来源于Mapillary平台并遵循CC-BY-SA 4.0许可。数据集包含多个城市的GEOJSON文件，文件大小各异，适用于计算机视觉和城市信息学等领域的研究。