Super-resolution Digital Bathymetric Model Reconstruction: A Benchmark Dataset and A Comparison of Deep learning Models

This is a dataset specifically collected for super-resolution digital bathymetric model (DBM) reconstruction. This dataset contains 26,443 low-resolution-high-resolution DBM pairs, with 80% (21,155 pairs) allocated for training and 20% (5,288 pairs) for testing. The low-resolution DBMs have a resolution of 15 arcseconds, and the high-resolution DBMs have resolutions of 5, 3 and 1 arcseconds, respectively. This dataset supports research on various upsampling factors ranging from 15× to 3×. All DBMs have been simultaneously normalized using maximum-minimum values, with a minimum value of -10,790.0 and a maximum value of 6,590.0. Additionally, we provide the code for 15 models in TensorFlow platform.