Inter-Information Learning Enhances Rail Surface Defect Segmentation in Multi-Task Learning

These findings suggest that the Inter-Information Learning (IIL) in the Multi-Task Learning (MTL) module plays a crucial role in producing improved representations for the segmentation of rail surface defects. As anticipated, the IIL effectively facilitates the exchange of information between the edge sub-network and the segmentation sub-network within the MTL module, resulting in enhanced segmentation outcomes.