LSKSANet: A Novel Architecture for Remote Sensing Image Semantic Segmentation Leveraging Large Selective Kernel and Sparse Attention Mechanism

TL;DR

This paper introduces LSKSANet, a lightweight remote sensing image segmentation model that combines large selective kernels and sparse attention to improve accuracy while maintaining efficiency.

Contribution

The paper presents a novel architecture integrating large selective kernels with sparse attention, enhancing receptive field and feature aggregation for remote sensing segmentation.

Findings

Outperforms state-of-the-art methods on Vaihingen dataset

Achieves superior segmentation accuracy on Postdam dataset

Maintains computational efficiency with a lightweight design

Abstract

In this paper, we proposed large selective kernel and sparse attention network (LSKSANet) for remote sensing image semantic segmentation. The LSKSANet is a lightweight network that effectively combines convolution with sparse attention mechanisms. Specifically, we design large selective kernel module to decomposing the large kernel into a series of depth-wise convolutions with progressively increasing dilation rates, thereby expanding the receptive field without significantly increasing the computational burden. In addition, we introduce the sparse attention to keep the most useful self-attention values for better feature aggregation. Experimental results on the Vaihingen and Postdam datasets demonstrate the superior performance of the proposed LSKSANet over state-of-the-art methods.