Railway Sleeper Spacing UAV Dataset (RSS UAV Dataset)

The Railway Sleeper Spacing UAV Dataset (RSS UAV Dataset) is a UAV-based railway infrastructure dataset developed for railway sleeper and rail detection, segmentation, and geometric measurement applications. The dataset is designed to support computer vision tasks involving railway track inspection, sleeper localisation, segmentation refinement, and centre-to-centre sleeper spacing estimation under real field conditions. The dataset contains 1400 high resolution aerial images collected from operational railway corridors using a DJI Mavic 3 UAV platform. Images were acquired from the Mau–Aurnihar railway section within the Varanasi Division of Indian Railways under varying environmental and illumination conditions, including open track sections, yard regions, bridge approach sections, ballast variability, and partial shadow conditions. Dataset Characteristics Total Images: 1400 Image Resolution: 3840 × 2160 Total Annotations: 51,398 Rail Annotations: 5,597 Sleeper Annotations: 45,801 Annotation Classes: Rail and Sleeper Annotation Format: YOLO annotation format Track Configuration: Multi-track railway Sleeper Types: Two concrete sleeper types Dataset Split Training Set: 980 images Validation Set: 280 images Test Set: 140 images Applications The dataset is intended for applications involving railway sleeper detection, segmentation, geometric measurement, UAV-based railway inspection, and spacing estimation research. The dataset can also support benchmarking of object detection and segmentation methods under repetitive infrastructure environments.