MoSeVa: model selection with spectral variability based on manifold learning.

This is the repo for MoSeVa: an automatic identification and quantification algorithm which considers spectral variability for gamma-ray spectrometry.

The code is organized as follows:

• The Code folder contains the source code for the IAE and the MoSeVa ,P-OMP algorithm
• The Data folder contains the dataset of 96 spectral signatures of 12 radionuclides as a function of steel thickness.
• The Notebooks folder contains two jupyter notebook files for training an IAE model and using MoSeVa to identify and quantify the radionuclides
  • The Models folder contains the pre-trained IAE model.

Package requirements

MoSeVa was coded using Pytorch. To use MoSeVa, you will need the packages listed in environment.yml. To create and activate a conda environment with all the imports needed, do (with CPU):

• conda env create -f environment.yml
• conda activate pytorch

If there is a problem with the installation of Pytorch, please follow this link to install it correctly: Pytorch.

Test MoSeVa code

• Run IAE_CNN_joint_gamma_spectrometry.ipynb if you want to see how IAE works and train a new IAE model with your new data.
• Run Identification_quantification_variability.ipynb if you want to see how to use MoSeVa.

Test on another data

• Replace the data in the Data folder with your own data
• If you want to use P-OMP (without variabilty):
  • Simply change the data in the file Identification_quantification_variability.ipynb
• If you want to use MoSeVa (with variabilty):
  • Train a new IAE model in IAE_CNN_joint_gamma_spectrometry.ipynb.
  • Here are a few tips on how to train IAE effectively: increase the number of anchor points if the variability is very complex, increase the number of layers if the result is not very good.
  • Update the new IAE model in Identification_quantification_variability.ipynb. The MoSeVa code normally works well with predefined parameters in the notebook.