Effect of non-resonant background on the extraction of Raman signals from CARS spectra using deep neural networks
By Rajendhar junjuri, Ali Saghi, Lasse Lensu, and Erik M. Vartiainen
Three different types of non-resonant backgrounds (NRBs) were explored to simulate the CARS spectra viz (1) product of two sigmoids following the original SpecNet model, (2) Single Sigmoid, and (3) fourth-order polynomial function.
DOI: https://doi.org/10.1039/D2RA03983D
Junjuri, R., Saghi, A., Lensu, L., & Vartiainen, E. M. (2022). Effect of non-resonant background on the extraction of Raman signals from CARS spectra using deep neural networks. RSC advances, 12(44), 28755-28766.
https://scholar.google.co.in/citations?hl=en&user=BRu_wuAAAAAJ&view_op=list_works&sortby=pubdate
These are 300 synthetic test spectra evaluated in the manuscript.
-
First, 100 spectra correspond to ‘Product of two Sigmoid NRB’.
-
Spectra 101-200 account for ‘One sigmoid NRB’.
-
Spectra 201-300 correspond to ‘Polynomial NRB’.
"y_test_300_merge_spectra3.npy"---> referes to the true Raman signal
"x_test_300_merge_spectra3.npy"---> referes to the input CARS data
The experimental CARS test data set used in this investigation can only be provided upon request and can contact Erik M. Vartiainen
The model architecture is directly adapted from the SpecNet paper (See https://github.com/Valensicv/SpecNet for the full code of the neural network model) Here three different NRBs are evaluated.
It can be accessed from the following program. RSS_Advances_CNN_to_train_with_different_NRBs.py
Testing can be done by using the following program. RSS_Advances_CNN_prediction_on_test_data.py
"One_sigmoid_NRB_model_weights.h5" --->referes weights of the model trained with One sigmoid NRB.
"Polynomial_NRB_model_weights.h5" --->referes weights of the model trained with Polynomial_NRB.
"Specnet_weights.h5" --->referes weights of the model trained with the product of two sigmoids NRB
You can use Python (TensorFlow 2.7.0) to test the pre-trained network. We have tested it in Spyder.