reviseDeep learning technology has been widely used in rail surface defect segmentation because of its strong feature representation ability However many existing methods based on deep learning often

Deep learning technology has gained significant popularity in rail surface defect segmentation due to its exceptional ability to represent features. Nevertheless, many current deep learning methods fall short in accurately delineating boundaries and delivering satisfactory defect region segmentation results. This limitation primarily arises from inadequate integration of contextual information and insufficient differentiation of boundary features.