GASS-Method/SVI4BuildingFunc

该数据集涵盖了四个城市，其中三个位于美国：纽约、洛杉矶、波士顿，一个位于西班牙：马德里。纽约的数据改编自OmniCity数据集，保留了其训练集中的3,100张标记图像和测试集中的3,300张标记图像，并补充了31,000张未标记图像用于半监督训练。马德里的数据包括训练集中的1,000张标记图像和测试集中的2,000张标记图像，以及额外的10,000张未标记图像。为了支持不同分类系统的研究，数据集进一步扩展到了洛杉矶和波士顿。

The SVI4BuildingFunc dataset is a street view image dataset that includes images from four cities, along with building function annotations based on the classification system of each city. The dataset covers three cities in the United States (New York, Los Angeles, Boston) and one city in Spain (Madrid). The data for New York is adapted from the OmniCity dataset, retaining 3,100 labeled training images and 3,300 labeled test images, supplemented with 31,000 unlabeled images for semi-supervised training. The Madrid data includes 1,000 labeled training images, 2,000 labeled test images, and an additional 10,000 unlabeled images. To support research on different categorization systems, the dataset was further expanded to include Los Angeles and Boston.