"Dataset of Optimal Radio Maps and Deployments"

"This dataset provides comprehensive optimal transmitter placement solutions for 167,525 urban building scenarios, generated using deep learning-based radio propagation prediction. For each building, the dataset includes: (1) predicted received power and coverage maps for the optimal transmitter location under both coverage-maximization and average received power objectives, (2) spatially-averaged power and coverage maps across all candidate locations, and (3) ranked lists of the top-500 transmitter locations optimized for coverage and power separately.The dataset was generated using SAIPP-Net, a convolutional neural network trained on ray-tracing simulations, to predict path loss for 256\u00d7256 meter urban areas. Each building's center 150\u00d7150 meter region was exhaustively evaluated, with predictions stored as both 8-bit images (for best transmitter visualizations) and 16-bit images (for high-precision averaged maps). Rankings account for coverage count (number of pixels receiving signal), average signal power, and total signal power in the analysis region."