A novel semi-supervised approach for semantic segmentation of aerial remote sensing images under limited ground-truth availability

Abstract Conventional semantic segmentation techniques rely heavily on the availability of substantial ground-truth data. However, this prerequisite often proves infeasible in real-world scenarios, particularly withthe labeling complexities inherent in remote sensing images. In this manuscript, a semi-supervisedapproach has been investigated towards semantic segmentation of remotely sensed images by address-ing the challenge of limited availability of ground-truth information. For this purpose, a hybridintegration of a standard semantic segmentation model and an adversarial model has been proposedunder semi-supervised setting. The former predict the masks for the unlabelled images when fine-tunedwith the available labelled training images (however limited they may be); whereas the latter aids thereconstruction of original input images from the predicted soft(masks) through an adversarial mecha-nism. This reconstruction, further validated through a reconstruction score, assist in the identificationof ‘most-confident’ image-mask pairs to be strategically integrated into the training set. The contri-bution ultimately is to utilise the unannotated images to meaningfully augment the limited trainingset to obtain an enhanced one. The proposed technique showcases a significant improvement, withan 11-34% enhancement over existing approaches in terms of mean intersection over union, precision,and F1-score across both the minifrance (MF) and dense labeling remote sensing dataset (DLRSD) datasets.