Deep Learning Approach to Line of Sight Detection in Urban Environments

This work proposes a deep learning-based approach for accurate Line-of-Sight prediction in complex urban environments. Leveraging a U-Net convolutional neural network, the model directly processes image-based city maps to predict LoS coverage maps. Ground truth images, necessary for both training and testing, are generated through ray-tracing simulations. Unlike traditional statistical or tabular machine learning models, this method automatically learns spatial features and dependencies without requiring manual feature engineering. Results demonstrate high accuracy, precision, recall, and F1score, confirming the potential of convolutional architectures in wireless channel modeling tasks. The proposed framework offers a scalable and real-time solution for LoS detection, enabling more efficient and adaptive wireless system deployment.