Mining self-similarity: Label super-resolution with epitomic representations

The subfolder code/data includes an example of a 2×downsampled 2×2km tile of aerial imagery and the associated high-res and low-res labels. Run the notebook trainepitome.ipynb to visualize the iterations of epitome training on this area. Run selfepitomicsr.ipynb to visualize single-image label super-resolution. The corresponding static HTML files can also be used for viewing.