Data and code for deep learning denoising with a residual U-Net for accelerated computation of three-dimensional Monte-Carlo radiation fields

This dataset supports the study entitled “Deep learning denoising with a residual U-Net for accelerated computation of three-dimensional Monte-Carlo radiation fields.” It contains paired three-dimensional Monte Carlo radiation fields for two simulation cases: a photon waste-drum room and a neutron nested-shielding room. For each case, 100 source, geometry, or shielding configurations were generated through random parameter sampling.

Monte Carlo simulations with 10^6 and 10^9 particle histories were performed for each configuration to produce paired low-particle-count noisy fields and high-particle-count reference fields. The samples were divided into 70 training, 15 validation, and 15 test pairs.For the photon case, dose-rate fields were tallied on a 200 × 50 × 70 grid with a voxel resolution of 5 cm. The varied parameters include the number, positions, densities, activities, and energy spectra of the waste drums.