This article presents an approach based on Generative Adversarial Neural Networks for the super-resolution problem of a Digital Elevation Model. A customized version of the Super-Resolution Generative Adversarial Network (SRGAN), referred to as SRGAN C, was developed to improve DEM quality and performance metrics compared to the original SRGAN, as well as to bilinear and bicubic interpolation methods. The evaluation focused on key metrics to assess the quality of the generated super-resolution data. The model was validated using real DEM data from different regions of Uruguay, gathering diverse terrain features to enhance the generalization capability of the trained models, enabling their potential application to the entire country. The obtained results show that significant improvements were computed in quantitative and qualitative metrics compared to traditional interpolation methods commonly used for the problem.