Abstract: In the task of semantic understanding of remote sensing images, most current research focuses on learning contextual information through attention mechanisms or multiple inductive biases.