Introduction：

The Multi-modal Remote Sensing Scene Classification Dataset (MRSSC2.0) is a cross domain remote sensing scene classification dataset with four modes of Visible Near Infrared(VIS), Short Wavelength Infrared(SWI), Thermal Infrared(INF) and Synthetic Aperture Radar(SAR), which is constructed based on the high-quality earth observation data obtained by Tiangong-2 Wide-band Imaging Spectrometer and Interferometric Imaging Radar Altimeter. The dataset contains 26710 images of 7 typical scenes including city, farmland, mountain, desert, coast, lake and river. In addition, MRSSC2.0 provides an extended task: weakly supervised cross domain target detection. MRSSC2.0 will promote the development of cross domain remote sensing scene classification and stimulate innovative research in weak surveillance target detection and domain adaptive data retrieval.

Sample Images：

Downloads：

Baidu netdisk:

https://pan.baidu.com/s/1WFxqBlnxsQa0ddS6V3ESbQ  (password: m64g)

Google drive:

https://drive.google.com/drive/folders/1p5oAVuWqyf81I03Ue1m7NyM7P7vrLDZn?usp=sharing

Publications:

Liu K, Yang J, Li S. Remote-Sensing Cross-Domain Scene Classification: A Dataset and Benchmark[J]. Remote Sensing, 2022, 14(18): 4635.