Abstract:
Artificial Intelligence for digital REStoration of Cultural Heritage (AIRES-CH) is a project focused on a creation of a cloud-native web app for the digital restoration of pictorial artworks through Computer Vision technologies applied to nuclear imaging raw data, such as X-Ray Fluorescence (XRF). In this talk, after a really brief introduction on nuclear imaging technologies, I will show that the task of associating an RGB colour image to an XRF imaging raw data is feasible by means of a multidimensional neural network, by discussing the Deep Neural Network models pipeline (training dataset composition, models’ architecture definition, training process, hyperparameter optimisation, validation); I will close with a brief discussion on how the trained neural network is employed in the cloud-native web application for XRF raw data real-time analysis.