RadioMap-PCE: A Multiscale Radio Map Dataset for Physical Coverage Extrapolation

RadioMap-PCE is a simulated radio-map dataset for physical coverage extrapolation at a fixed spatial resolution of 1 m/pixel. It contains 1,400 urban scenes and 28,000 map–transmitter samples across five native square grids with side lengths of 256, 384, 512, 768, and 1024 pixels. The 256-scale subset contains 1,000 scenes and 20,000 samples; each larger-scale subset contains 100 scenes and 2,000 samples. Each sample provides a building occupancy image, a single-transmitter location image, and an 8-bit normalized channel-gain label generated using the Dominant Path Model in Altair WinProp. Shared settings include a carrier frequency of 5.9 GHz, transmit power of 23 dBm, transmitter and receiver heights of 1.5 m, and building height of 25 m. Building geometries originate from OpenStreetMap. The release includes processed building polygons, transmitter coordinates, portable sample and scene indexes, simulation metadata, and exact map-wise benchmark splits. The scales are not spatially paired. The main test protocol contains 90 scenes and 1,800 samples at each scale. Separate metadata document the historical 768-scale zero-shot baseline subset and restoration-control validation subsets. The accompanying README explains label conversion, directory layout, benchmark label access, and a minimal Python loading example. This dataset accompanies the manuscript “ScaleProp: Label-Efficient Radio Map Prediction for Physical Coverage Extrapolation”. Dataset creator and contact: Tianqi Mao, [e-mail]. The ScaleProp implementation is planned for public release after publication of the associated article. The original contributions are licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). Building data contain information from © OpenStreetMap contributors, available under the Open Database License (ODbL) 1.0. See the included LICENSE.md and THIRD_PARTY_NOTICES.md for component-specific license scope and attribution.