TurnoutSegmentation: A Semantic Segmentation Dataset for Railway Single Turnouts

The TurnoutSegmentation dataset is the first publicly available dataset for fine-grained visual analysis of railway switch equipment, providing key data support to address bottlenecks such as the lack of visualization of mechanical structural deformation and insufficient modeling of electrical mechanical state associations in traditional detection. Its core value lies in providing standardized pixel level semantic labels for components (sharp rails, switch machines, basic rails, and close fitting adjustment rods), directly serving intelligent operation and maintenance scenarios: achieving component gap quantification through precise segmentation, supporting cross domain analysis such as track state evaluation and predictive maintenance. The data aggregation is designed as a ready to use benchmark, consisting of 2360 high-definition images and accompanying annotations, divided into a training set (1610 frames), a validation set (691 frames), and a test set (59 frames). Users can use component masks to develop defect detection algorithms, or combine position coordinates to achieve deformation displacement analysis.