Dataset of cycling-specific infrastructure in Switzerland (1998-2024)

The dataset contains 616,948 georeferenced instances of cycling-specific infrastructure features across 10 major Swiss agglomerations (Basel, Bern, Biel, Geneva, Lausanne, Lugano, Lucerne, St. Gallen, Winterthur, Zurich), extracted from aerial imagery (SWISSIMAGE) between 1998 and 2023. Model and codebase available: https://gitlab.ethz.ch/spapini/remote_velo_public Each record includes: instance_id: Unique identifier for each detected object. class_type: One of 11 infrastructure types: advanced_retaining_bar bike_lane bike_pictogram bike_stopping_space circle (roundabouts) crosswalk pedestrian_island red_colored_bike_lane retaining_bar_white / retaining_bar_yellow shark_teeth_white / shark_teeth_yellow year: Year the infrastructure was detected. year_first / year_last: First and last year the object was detected (used to infer persistence). mean_confidence: Average model confidence score for the detection. n: Number of times the object was detected across years (post-processed). resolution: Ground resolution of the aerial image (10, 25, or 50 cm). agglomeration: Name of the urban area. bfs_gemein_id / gemeinde_name / kanton_id: Administrative identifiers (municipality and canton). geometry: Spatial polygon (in CH1903+ / LV95 coordinate system) representing the object’s location and shape.