Replication Data for: Análise Comparativa de Redes Neurais Convolucionais para Super-Resolução de Imagens Naturais e Sísmicas

It comprises trained weights for five Fully Convolutional Networks for Super-Resolution of photographic and seismic images at 2 and 4 times scale factors. Each network was trained five times on each task, totaling 100 trained models.

It also contains per-image PSNR scores for each model on the DIV2K, Unicamp-NAMSS, and Urban100 datasets. A Jupyter Notebook is provided to compute the partial ordering presented on the results chapter, from the raw PSNR tables. See README.md for details.