Artificial Intelligence for pigment classification task in painted works of art

Webinar on 2^nd International Congress on AI and Machine Learning

February 15, 2022 | Webinar

Tsveta Miteva

Sorbonne Universit�?©, France

ScientificTracks Abstracts: RRJET

Abstract

Hyperspectral reflectance imaging in the short-wave infrared range (SWIR, ca. 1000 to 2500 nm) has proven to provide enhanced characterization of paint materials. However, the interpretation of the results remains challenging due to the intrinsic complexity of the SWIR spectra, presenting both broad and narrow absorption features with possible overlaps. To cope with the high dimensionality and spectral complexity of such datasets acquired in the SWIR domain, a promising data treatment approach is presented, inspired by innovative development in the cultural heritage field: the use of a pigment spectral database (extracted from model and historical samples) combined with a deep neural network (DNN). This approach allows for multi-label pigment classification within each pixel of the hyperspectral data cube. The DNN results are compared with Conventional Spectral Angle Mapping and the superior classification ability of the DNN is outlined. DNN results obtained on both pigment reference samples and a Buddhist painting (thangka) are discussed.

Biography

Dr Tsveta Miteva has completed her PhD from the University of Heidelberg, Germany, and postdoctoral studies from Sorbonne University in Paris, France. Since 2018, Dr Miteva is a researcher at the French National Scientific Center (CNRS), working at Laboratoire de Chimie Physique—Matière et Rayonnement. Her research focuses on the theoretical simulation of ultrafast electronic decay processes following inner-shell excitation and ionization in gas and liquid phases, and on applicatons of Machine Learning and Artificial Intelligence. She has published 34 papers in renowned journals..

• PDF
• HTML