Dense Fusion Classmate Network for Land Cover Classification

Abstract

Recently, FCNs based methods have made great progress in semantic segmentation. Different with ordinary scenes, satellite image owns specific characteristics, which elements always extend to large scope and no regular or clear boundaries. Therefore, effective mid-level structure information extremely missing, precise pixel-level classification becomes tough issues. In this paper, a Dense Fusion Classmate Network (DFCNet) is proposed to adopt in land cover classification. DFCNet is jointly trained with auxiliary road dataset seemed as “classmate”, which properly compensates the lack of mid-level information. Meanwhile, a dense fusion module is also integrated, which guarantees the precise discrimination of confused pixels and benefits the network optimization from scratch. Score on Deep- Globe land cover classification competition shows that our approach has achieved good performance.