Dayton is a dataset for ground-to-air (or air-to-ground) image conversion or cross-view image synthesis. The dataset contains 76,048 images of road view and road bird's-eye view (training/testing split is 55,000/21,048). The image resolution in the original dataset is 354×354.