reviseThe deep learning technology has been widely applied to segment rail surface defects due to the powerful feature representation ability However most of the existing deep learning-based meth-ods

Deep learning technology has found extensive applications in segmenting rail surface defects, owing to its strong feature representation capability. Nevertheless, many existing methods based on deep learning tend to yield imprecise boundaries for defect regions and unsatisfactory segmentation outcomes, primarily due to the insufficient context information and inadequate discriminative features.