Semantic Segmentation on Remotely-Sensed Images Using Enhanced Global Convolutional Network with Channel Attention and Domain Specific Transfer Learning

In remote sensing domain, it is crucial to automatically annotate semantics, e.g., river, building, forest, etc, on the raster images. Deep Convolutional Encoder Decoder (DCED) network is the state-of-the-art semantic segmentation for remotely-sensed images. However, the accuracy is still limited, since the network is not designed for remotely sensed images and the training data in this domain is deficient. In this paper, we aim to propose a novel CNN network for semantic segmentation particularly for remote sensing corpora with three main contributions. First, we propose to apply a recent CNN network call ''Global Convolutional Network (GCN)'', since it can capture different resolutions by extracting multi-scale features from different stages of the network. Also, we further enhance the network by improving its backbone using larger numbers of layers, which is suitable for medium resolution remotely sensed images. Second, ''Channel Attention'' is presented into our network in order to select most discriminative filters (features). Third, ''Domain Specific Transfer Learning'' is introduced to alleviate the scarcity issue by utilizing other remotely sensed corpora with different resolutions as pre-trained data. The experiment was then conducted on two given data sets: ($i$) medium resolution data collected from Landsat-8 satellite and ($ii$) very high resolution data called ''ISPRS Vaihingen Challenge Data Set''. The results show that our networks outperformed DCED in terms of $F1$ for 17.48% and 2.49% on medium and very high resolution corpora, respectively.