MLRaman

Rapid Machine Learning–Driven Detection of Pesticides and Dyes Using Raman Spectroscopy

Requirement

MLRaman requires the packages as follows:

• torch: an open-source machine learning library with strong GPU acceleration.
• jupyterlab: a web-based interactive development environment for notebooks, code, and data.
• matplotlib: a comprehensive library for creating static, animated, and interactive visualizations in Python.
• seaborn: a Python data visualization library based on matplotlib.
• scikit-learn: a set of python modules for machine learning and data mining
• python-docx: a Python library for creating and updating Microsoft Word (.docx) files
• imagehash: a Python library for image hashing.
• opencv-python: a computer vision library, containing over 2500 algorithms, and is free for commercial use.
• xgboost: a machine learning library based on the gradient boosted decision trees algorithm.
• openTSNE: a modular Python implementation of t-Distributed Stochasitc Neighbor Embedding (t-SNE).
• umap-learn: a dimension reduction technique that can be used for the UMAP method.
• streamlit: a Python framework for data scientists and AI/ML engineers to deliver interactive data apps.

Example to install requirements with conda for CPU & GPU:

$ conda create -n torch python=3.9
$ conda activate torch
$ pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
$ pip3 install jupyterlab matplotlib seaborn scikit-learn python-docx imagehash opencv-python xgboost openTSNE umap-learn streamlit

Directory Description

MLRaman
├── data
│   └── docx: a dataset of Raman spectra of 10 pesticides and dyes, which humans collect from the literature.
│   └── test: the unseen experimental Raman spectra of CV and R6B for testing the app.
├── utils
│   ├── utils_data.py: define the function related to the dataset.
│   ├── utils_model.py: define the function related to the model.
│   └── utils_plot.py: define the function related to plotting results.
├── model
│   ├── resnet18_10cls_20251115.pth: a trained GNN model.
├── output
│   ├── xgb_on_cnn.pkl: a trained XGBoost model with GNN features and PCA.
├── MLRaman.ipynb: Main MLRaman code
└── app_xgb_streamlit.py: user-friendly Streamlit application for real-time prediction.

How to run

Step 1: Download the MLRaman package:

git clone https://github.com/nguyen-group/GNNOpt.git

Step 2: Go to the source code in the Raman directory to run the program:

cd MLRaman
jupyter-lab MLRaman.ipynb

Step 3: For Streamlit application:

streamlit run app_xgb_streamlit.py

Note: app_xgb_streamlit.py will load resnet18_10cls_20251115.pth and xgb_on_cnn.pkl, which are stored in the model and output directories, respectively.

References and citing

The detailed MLRaman is described in our pre-print:

TBA