ARQGAN - ARCHER Classical-period Doric Greek temples - obstacles dataset

This dataset has been used to develop the ARQGAN-ARCHER project (https://sites.google.com/view/arqgan), a research initiative focused on the reconstruction of architectural heritage through artificial intelligence (funded by: MICIU/AEI FEDER/UE, PID2021-126633NA-I00). It consists of a collection of images representing various examples of Classical Greek temples, generated from 3D models. This research has been carried out in collaboration with architects and engineers from the Polytechnic School at Universidad Francisco de Vitoria. Furthermore, this set of synthetic images has also been employed for the Master’s thesis of Iván Barcia, entitled “Reconstruction of Greek Temples with Suppression of Intrusive Elements Using GAN Networks”. The thesis was supervised by Professor Antonio Jesús Fernández García within the Master’s Degree in Artificial Intelligence at the International University of La Rioja (UNIR). We have used these images in the following articles. Please cite them if you make use of the dataset. https://doi.org/10.1016/j.eswa.2021.11509 https://doi.org/10.3390/app14166854 https://doi.org/10.1007/978-3-031-62963-1_51 https://doi.org/10.1007/978-3-031-36155-5_26 This dataset contains images of 30 Classical Greek temples, generated from 3D models. Some of them are theoretical, but they conform to the requirements of the Doric order. Each case study provides a set of 360 images (1024×1024 px), along with three additional sets of images representing different states of ruin. In total, each case study comprises 1,440 images. The overall dataset contains 43,200 images. The images of ruined buildings include obstacles such as trees, people, or rocks, which were used to train a neural network (GAN) designed to identify these intrusions and automatically reconstruct the complete building.