Dual-Task Semantic Change Detection for Remote Sensing Images Using the Generative Change Field Module

With the advent of very-high-resolution remote sensing images, semantic change detection (SCD) based on deep learning has become a research hotspot in recent years. SCD aims to observe Earth’s land surface and plays vital role monitoring ecological environment, use cover. Existing mainly focus single-task detection; problem they face is that existing methods are incapable identifying which type occurred each multi-temporal image. In addition, few binary region help train SCD-based network. Hence, we propose dual-task network (GCF-SCD-Net) by using generative field (GCF) module locate segment region; what more, proposed end-to-end trainable. meantime, because influence imbalance label, separable loss function alleviate over-fitting problem. Extensive experiments conducted this work validate performance our method. Finally, achieves 69.9% mIoU 17.9 Sek SECOND dataset. Compared with traditional networks, GCF-SCD-Net best results promising performances.