This paper aims to present a wide dataset of ground penetrating radar (GPR) profiles recorded on a full-size geophysical test site, in Nantes (France). The geophysical test site was conceived to reproduce objects and obstacles commonly met in the urban subsurface, in a completely controlled environment. Since the design phase, the site was especially adapted to the context of radar-based techniques. The main needs for urban civil engineering, in which the geophysical test site intends to provide an experimental field, relate to the detection, localization and characterization of hidden structures and not to the assessment of soil properties. As a tool accessible to professionals, the site has been used for various needs over the years, such as tests or comparisons of devices, and validation of measurement or modelling methods. After a detailled description of the site and its building process, the GPR profiles included in the dataset are reported and commented on. Overall, the dataset is composed by 67 profiles. They were recorded along eleven parallel lines, crossing the test site in the transverse direction, by using three pulsed radar systems, manufactured by different producers and equipped with various antennas having central frequencies from 200 MHz to 900 MHz. An archive containing all profiles (raw data) is enclosed to this paper as supplementary material. The dataset is the core part of the Open Database of Radargrams initiative of COST (European Cooperation in Science and Technology) Action TU1208 "Civil engineering applications of Ground Penetrating Radar". The idea beyond such initiative is to share with the scientific community a selection of interesting and reliable GPR responses, to enable an effective benchmark for direct and inverse electromagnetic approaches, as well as for imaging and signal processing algorithms. We hope that the dataset presented in this paper will be enriched by the contributions of further users in the future, who will visit the test site and acquire new data with their GPR systems. Moreover, we hope that the dataset will be made alive by researchers who will perform advanced analyses of the profiles, measure the electromagnetic characteristics of the host materials, contribute with synthetic radargrams obtained by modelling the site with electromagnetic simulators, and more in general share results achieved by applying their techniques on the available profiles.