The repo provide 2 tools:  

• FRDA, a data augmentation tool for FTIR samples

• 1D-CAM, a Explainable AI tool for analysing FTIR samples.

Here is the introduction of each code file.

• FTIR_1DCAM.py is used for training the 1DCAM model that is built by Keras. You can use it to train you own model,
• FTIR_AugmentationBased....py is used for ESMA methods. 
• FTIR_GaussinMixture.py is used for clustering each influence curve obtained from the 1DCAM and returning the separate point.
• FTIR_XCNN is the identical file to the FTIR_1DCAM.
• FTIR_fit_least_square.py is used for separating the curve and randomly selecting and combining the data.
• PLS.py is used for baseline reducing algorithms.
• utils.py is used for reading data and plotting the confusion matrix.

This repo provides 3 datasets:

• Kedzierski’s

• Jung's

• Hannah's

The FRDA tool performs data augmentation for each dataset and then train and validate different ML models

Kedzierski’s dataset with FRDA

python FTIR_DataAugmentation_Kedzierski.py

Jung’s dataset with FRDA

python FTIR_DataAugmentation_Jung.py

Jung’s dataset with FRDA

python FTIR_DataAugmentation_Hannah.py

The 1D-CAM can be used as follows:

• Training a classification model using 1D-CNN layer
• Creating 1D-CAM based the trained network (using the layer before the output layer and parameters).
• Inputting the new sample data to classification model and obtaining classification result.
• Inputting the identical data into 1DCAM and using the classification result to calculate the influence curve.

Generating influence data curve using 1DCAM

python FTIR_1DCAM.py

Use the below bibtex to cite us.

@article{yan2023 FRDA,
  title={FRDA: Fingerprint Region based Data Augmentation using Explainable AI for FTIR based Microplastic Classification},
  author={Yan, Xinyu and Cao, Zhi and Murphy, Alan and Ye, Yuhang and Wang, Xinwu and Qiao, Yuansong},
  journal={},
  volume={},
  number={},
  pages={},
  year={2023},
  publisher={Elsevier}
}

@misc{yan2023ensemble,
  title={FRDA: Fingerprint Region based Data Augmentation using Explainable AI for FTIR based Microplastic Classification},
  author={Yan, Xinyu and Cao, Zhi and Murphy, Alan and Ye, Yuhang and Wang, Xinwu and Qiao, Yuansong},
  year={2023},
  publisher={Github},
  howpublished={\url{https://github.com/lyheiyu/Fingerprint-Region-based-Data-Augmentation-using-Explainable-AI-for-FTIR-based-MP-Classification/}},
}

Software Research Institute of Technological University of the Shannon: Midlands Midwest.