‘Nomadic’ adaptive analysis of heritage features through AI. Discovering Eladio Dieste's Cristo Obrero Church identities

The digital documentation of architectural and urban heritage for conservation purposes has been interested in recent years by process innovation aimed at the standardisation of analysis methods, through the use of artificial intelligence (AI) algorithms. Repetitive and quantitative analysis operations can be delegated to computers, which learn from input data trained by specialists (architects, conservators) and provide support for their extensive interpretation. The most used data are point clouds and images, which can be segmented and classified according to different thematic levels. Historical heritage, however, has extremely variable (intrinsic and boundary) characteristics, which make each case specific. This implies, in many situations, that automatic and semiautomatic procedures, like ‘nomads’, are able to adapt to each building to which they are applied, through a multiscale approach, considering the specific characteristics of each one. This contribution, therefore, presents a method of analysing the state of conservation of a building in a Uruguayan urban suburb (Cristo Obrero Church, designed by Eladio Dieste in 1956) through the application of supervised Machine Learning algorithms.