Automatic Virtual Reconstruction of Historic Buildings Through Deep Learning. A Critical Analysis of a Paradigm Shift

New advances in the field of artificial intelligence propose to rethink methodologies in many areas of knowledge. In the field of Architectural Heritage studies, the virtual reconstruction of historical buildings in ruins has maintained the same analysis methodology for centuries. New technologies have been adding tools to the reconstruction process, which still depends on the theoretical assumption of a specialist. However, the development of neural networks (Deep Learning) is proposing a radical change in the analysis and reconstruction methodology. The proposal described as automatic virtual reconstruction uses Generative Adversarial Networks (GAN) and Natural Processing Language (NPL) technology for training networks and learning patterns of a specific architectural style. A comparative study of the two methods can enlighten on the possibilities that will emerge in the coming years. This paradigm shift in the virtual reconstruction of historic buildings can revolutionize, from a scientific and informative point of view, the way of understanding and interpreting architectural heritage.