Singapore dataset

The goal to create the Singapore dataset is to provide the worldwide researchers of remote sensing community a database for training and evaluating their algorithms.

Singapore dataset contains 1086 images, which was constructed from Google Earth, of Singapore. All images are stored in color tiff format with a resolution of 256*256 pixels. The spatial resolution per pixel is about 0.6 m. The dataset contains 9 challenging scene categories: airplane, forest, harbor, industry, meadow, overpass, residential, river and runway. Some typical images are displayed as below.

Class 1: airplane

Class 2: forest

Class 3: harbor

Class 4: industry

Class 5: meadow

Class 6: overpass

Class 7: residential

Class 8: river

Class 9: runway

Download notice:
Download the database ( rar file ), and please send an email to Qingyong Li (

If you use this database, please cite the following publication:
Jinrui Gan, Qingyong Li, Zhen Zhang et al. Two-Level Feature Representation for Aerial Scene Classification [J]. IEEE Geoscience and Remote Sensing Letters, 2016, 13(11): 1626-1630.