Remote Sensing Image Translation via Style-Based Recalibration Module and Improved Style Discriminator

Existing remote sensing change detection methods are heavily affected by seasonal variation. Since vegetation colors different between winter and summer, such variations inclined to be falsely detected as changes. In this letter, we proposed an image translation method solve the problem. A style-based recalibration module is introduced capture features effectively. Then, a new style discriminator designed improve performance. The can not only produce decision for fake or real sample but also return vector according channel-wise correlations. Extensive experiments conducted on season-varying data set. experimental results show that effectively perform translation, thereby consistently improving Our codes available at https://github.com/summitgao/RSIT_SRM_ISD .