Additional file 1 of An alternative approach to mapping pigments in paintings with hyperspectral reflectance image cubes using artificial intelligence

Additional file 1: Figure S1. The four paintings from the Laudario used to create the training dataset for the 1D-CNN model. a The Nativity with the Annunciation to the Shepherds, Master of the Dominican Effigies, c. 1340, miniature on vellum, National Gallery of Art, Washington, D.C., Rosenwald Collection, 1949.5.87, b The Ascension of Christ, Pacino di Bonaguida, about 1340, Tempera and gold leaf on parchment. Getty Museum, Los Angeles, Ms. 80a (2005.26), verso, c The Martyrdom of Saint Lawrence, Pacino di Bonaguida, about 1340, Tempera and gold leaf on parchment. Getty Museum, Los Angeles, Ms. 80b(2006.13), verso ( d) Christ and the Virgin Enthroned with Forty Saints, Master of the Dominican Effigies, c.1340, miniature on vellum, National Gallery of Art, Washington, D.C., Rosenwald Collection, 1959.16.2. Digital images of (b), (c). Table S1. Summary of per class accuracy for the 1D-CNN model. The 25 pigment/mixture classes and their total class size is given along with the per class accuracy using 10-fold cross validation.) courtesy of the Getty's Open Content Program. Figure S2. The average reectance spectra for each of the pigment labeled classes in the training dataset used for the 1D-CNN model. Figure S3. The pigment labeled map created by using a lower threshold (0.85) in the 1D-CNN. Table S2. Summary of analyses, the Pentecost. Note that the RIS features listed are those identified by an expert user following manual data exploration. When available, fiber optic reflectance spectroscopy (FORS) and Raman analysis may provide additional information about the total chemical composition of each area. However, not all pigments identified are discernible in the RIS data cube on which the 1D-CNN is applied. Therefore, a simplified “pigment class” column notes the materials that should be identified in the paint by this technique. Table S3. Summary of analyses, Saint Peter Enthroned. Note that the RIS features listed are those identified by an expert user following manual data exploration. When available, fiber optic reflectance spectroscopy (FORS) analysis may provide additional information about the total chemical composition of each area. However, not all pigments identified are discernible in the RIS data cube on which the 1D-CNN is applied. Therefore, a simplified “pigment class” column notes the materials that should be identified in the paint by this technique.