Skip to content

Physics-informed ML ODW

Physics-informed Machine Learning for Optically-Derived Water (ODW) prediction.

Overview

This package provides pre-trained machine learning models for predicting Optically-Derived Water (ODW) values from satellite imagery. It integrates with sensingpy for image processing workflows.

Features

  • Pre-trained Models: Ready-to-use ML pipelines for ODW prediction
  • Multiple Model Types:
    • ML: Standard Machine Learning model (MLP)
    • CS_ML: Caballero & Stumpf Machine Learning model
  • Easy Integration: Works seamlessly with sensingpy Image objects
  • 2D Prediction: Generate full raster predictions from satellite imagery

Installation

git clone https://github.com/Aouei/Physics-informed-ML-ODW.git
cd Physics-informed-ML-ODW
pip install .

Dependencies

  • Python >= 3.11
  • sensingpy
  • numpy
  • pandas
  • scikit-learn == 1.7.2

Quick Start

from sensingpy import reader
from physics_informed_ml_odw import predict_2d

# Load satellite image
image = reader.open('path/to/image.tif')

# Predict ODW values
image['ODW'] = predict_2d(image, model='ML')

Available Models

Model Description File
ML Machine Learning MLP model ML__MLP.pkl
CS_ML Caballero & Stumpf ML model CS_ML__MLP.pkl

License

This project is open source. See the repository for license details.