An Empirical Study on Using CNNs for Fast Radio Signal Prediction

Accurate radio frequency power prediction in a geographic region is computationally expensive part of finding the optimal transmitter location using ray tracing software. We empirically analyze viability deep learning models to speed up this process. Specifically, methods including CNNs and UNET are typically used for segmentation, can also be employed tasks. consider dataset that consists values five different regions with four frame dimensions. compare learning-based RadioUNET variations model task. More complex improve on higher resolution frames such as $$256\times 256$$ . However, same lower resolutions results overfitting simpler perform better. Our detailed numerical analysis shows effective they able generalize well new regions.