A generalized few-shot segmentation-based framework, named SegLand, is proposed to efficiently update novel land cover classes in high-resolution remote sensing imagery with limited labeled data.
