RoadMark-binary: 29,405 RGB aerial tiles with ground truth masks labeled with horizontal road line marking information

The dataset consists of 27,360 PNG images (256x256pixels) x 3 bands of high resolution aerial orthoimages at 0,15 m/pixel of resolution with a 12.5% overlap, covering a land area of 40.34 km². The images contain information related to representative road lines and markings found on highway pavement and is labelled for binary semantic segmentation. Test data: A novel area from the Madrid region labeled with binary road marking line information, divided into 2,045 RGB tiles of 256 × 256 pixels, with no tile overlap (land area of 3.02 km²). The “Road_Marking” class covers approximately 5% of the ground truth pixels. Road names included in dataset are: A-1101, A-129, A-1402, A-1506, A-2228, A-227, A-2302, A-331, A-333, A-4050, A-421, A-55, A-6_M-50, A-6200, AB-515, AC-400, AP-15AS-265, CL-602, CL-626, CL-631, CM-2105, CM-211, CM-3127, CM-3129, CM-403, CV-41, CV-70, CV-801, CV-845, DSA-469, EI-500, EX-104, EX-105, GC-2, GC-800, LZ-10M-50, M-510, Ma-1, Ma-12, Ma-15, MA-15_VDS1, Ma-19A, MA-3011, Ma-3220, Ma-3240, Ma-3301, Ma-3341, Ma-3510, Ma-5010, Ma-5020, Ma-5030, Ma-5040, Ma-5101N-322, N-340A, N-420-1, N-547, N-634, OU-302, RM-315, RM-E19, TF-64