Roads and bridges are essential to civilian, commercial, and government transport across the world. One facet of roads that is often ill-captured by GPS navigation systems is that of road overpasses, and this can have severe consequences for drivers of large vehicles. We aim to rectify this lack by predicting the presence of road overpasses in a given satellite image. To do so, we apply a convolutional neural network to a dataset of satellite images that we labeled and assembled.