The classifcaton of trace chemical residues through actve spectroscopic sensing is challenging due to the lack of physics-based models that can accurately predict spectra. To overcome this challenge, we leveraged the feld of domain adaptaton to translate data from the simulated to the measured domain for training a classifer. We developed the frst 1D conditonal generatve adversarial network (GAN) to perform spectrum-to-spectrum translaton of refectance signatures. We applied the 1D conditonal GAN to a library of simulated spectra and quantfed the improvement in classifcaton accuracy on real data using the translated spectra for training the classifer. Using the GAN-translated library, the average classifcaton accuracy increased from 0.622 to 0.723 on real chemical refectance data, including data from chemicals not included in the GAN training set.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
C.P. Murphy and J. Kerekes, “1D conditonal generatve adversarial network for spectrum-to-spectrum translaton of simulated chemical refectance signatures”, J. Spectral Imaging 10, a2 (2021). htps://doi.org/10.1255/jsi.2021.a2
RIT – Main Campus