Contrastive spectra matching

This repository provides the implementation for our paper Raman spectra matching with contrastive representation learning. We experimentally show that we significantly outperform or is on par with the existing approaches for Raman spectra identification on three publically available datasets.

Requirement

git clone https://github.com/lyn1874/raman_spectra_matching_with_contrastive_learning.git
cd contrastive_spectra_matching
conda env create -f spectra_matching.yaml
conda activate torch_dl

Testing

The top-1 matching process and conformal prediction process per dataset is shown in the jupyter file test_experiment.ipynb

Training

To train a spectra matching model for each dataset, run the following script:

./run_rruff.sh raw '0 1 2 3'
./run_rruff.sh excellent_unoriented '0 1 2 3'
./run_organic.sh raw '0 1 2 3'
./run_organic.sh preprocess '0 1 2 3'
./run_bacteria.sh bacteria_random_reference_finetune '0 1 2 3'

The repeat_g in each script represents the number of models that are used for the ensemble calculation

Reproduce figures

python paper_figures.py --index figure_augmentation_example --save False --pdf_pgf pdf

Dataset

You can download the Mineral and Organic dataset here: https://data.dtu.dk/articles/dataset/Datasets_for_replicating_the_paper_Raman_Spectrum_Matching_with_Contrastive_Representation_Learning_/20222331 As for the Bacteria dataset, you can visit the official repository https://github.com/csho33/bacteria-ID for downloading the dataset.

Experiments

You can download the experiments for each of the datasets here: https://data.dtu.dk/articles/dataset/Experiments_for_replicating_the_paper_Raman_spectrum_matching_with_contrastive_representation_learning_/21590934

Citation

If you use this code, please cite:

@Article{D2AN00403H,
author ="Li, Bo and Schmidt, Mikkel N. and Alstrøm, Tommy S.",
title  ="Raman spectrum matching with contrastive representation learning",
journal  ="Analyst",
year  ="2022",
volume  ="147",
issue  ="10",
pages  ="2238-2246",
publisher  ="The Royal Society of Chemistry",
doi  ="10.1039/D2AN00403H",
url  ="http://dx.doi.org/10.1039/D2AN00403H",
}