Remote Sensing Semantic Segmentation via Boundary Supervision-Aided Multiscale Channelwise Cross Attention Network

High spatial resolution (HSR) remote sensing (RS) images inevitably pose the challenge of multiscale transformation, as small objects, such as cars and helicopters (HCs), may occupy only a few pixel points. This incurs a significant hurdle for global context modeling, particularly in backbone networks with large downsampling coefficients. Simple summation or concatenation techniques, such as skip connections, fail to address semantic gaps and even impose negative impacts on multiscale feature fusion. Meanwhile, due to the complexity of foreground objects, the boundary details of HSR RS images are easy to lose in sampling operations. To overcome these challenges, we propose a multiscale channelwise cross attention network (MCCANet) assisted by boundary supervision (BS). Technically, MCCA captures the channel attention (CA) with various scales, which allows dynamic and adaptive feature fusion in a contextual scale-aware manner and focuses on both large and small objects distributed throughout the inputs. Besides, a channel and context strainer (CCS) module is proposed and embedded in MCCA, filtering channels and contexts for the mitigation of intraclass differences. In addition, we apply a BS module to recover boundary contour, avoiding the blurring effect during the construction of contextual information. The refined boundary allows for the effective recognition of surrounding pixels, ensuring a better segmentation performance. Extensive experiments on the instance segmentation in aerial images dataset (iSAID), International Society for Photogrammetry and Remote Sensing (ISPRS) Potsdam, and land-cover domain adaptive (LoveDA) datasets demonstrate that our proposed MCCANet achieves a good balance of high accuracy and efficiency. Code will be available at: https://github.com/ZhengJianwei2/MCCANet.