Unmixing ATR-{\mu}FTIR spectroscopic images of cross-sections of historical oil paintings
Researchers have developed a new unsupervised convolutional neural network autoencoder to analyze complex spectroscopic images from historical oil paintings. This method aims to automate the interpretation of Attenuated Total Reflection Fourier Transform Infrared Microscopy (ATR-μFTIR) hyperspectral images, which are currently analyzed manually. The new approach estimates spectral components and their distribution maps, incorporating a weighted spectral angle distance loss to improve accuracy and reduce sensitivity to acquisition artifacts. AI
IMPACT Automates complex analysis of historical artifacts, potentially accelerating art conservation research.