Physical prior-informed deep generative model for spectroscopy transfer and material characterization

We present SpectroGen, an advanced, physical prior-informed generative artificial intelligence model, to facilitate transformation between various spectroscopic techniques, revolutionizing spectral acquisition approaches. Our experimental results, demonstrating over 99% correlation and less than 0.01% root mean square error, validate that our approach can effectively mitigate the limitations of spectroscopic instruments, thereby advancing material characterization and biomedical diagnostics.