Current students


MILANI FEDERICOCycle: XXXV

Section: Computer Science and Engineering
Advisor: FRATERNALI PIERO
Tutor: GATTI NICOLA

Major Research topic:
Analysis of Cultural Heritage data for complex iconography studies and enrichment of online collections

Abstract:
;
;
With the massive digitalization of artworks and the increasing availability of open access online sources, researchers have the possibility to investigate thoroughly the history of art and the evolutions of motifs and themes across space and time. However, most available resources do not offer high-quality images and most data sets comprise low-quality or scarce annotations, that result in a substantial difficulty with respect to natural images. This is especially the case of iconography: the branch of history of art that studies the subject matter or meaning of works of art; it has a fundamental role as a cultural marker which can be used to support the identification of the production period and provenance of an artwork. Classifying artwork by their iconography has been little investigated and still presents non trivial challenges besides the scarsicity of data: iconographic classes are numerous and structurally complex  and the art images themselves are visually more homogeneous than natural images which makes their discrimination difficult. To deal with these problems, a combination of semi-supervised and weakly-supervised approaches can be exploited with the help of models trained on a large quantity of artworks or with the help of previous knowledge learnt from natural images. Novel Deep Learning techniques can be explored to train a model able to correctly identify iconographic elements in artworks along with their accurate positions and spatial relations while also dealing with the issues presented by noisy annotations and low-quality data. Ultimately, these results can be used to aid experts in their research with advanced studies, to provide novel augmented reality experiences to museums or to enrich online collections with metadata that otherwise would be too time-consuming to obtain.
;
;