This repository contains code for the following paper, under review in Springer Machine Learning:
p3VAE: a physics-integrated generative model. Application to the pixel-wise classification of hyperspectral images
A short version of this work has been accepted as a workshop paper at Machine Learning for Remote Sensing, ICLR 2024.
Please cite this paper if you use the code in this repository as part of a published research project (see bibtex citation below).
The code was run using python 3.8:
- create a python virtual environment
- clone this repo:
git clone https://github.com/Romain3Ch216/p3VAE.git - navigate to the repository:
cd p3VAE - install python requirements:
pip install -r requirements.txt
We provide the data and code that were used to compute results from experiments of section 5.
The train.py script was used to train the models which weights are in the results folder.
Other files were used to plot the figures of section 5.
For instance, to reproduce the figure 7 of section 5 for the p3VAE with seed 103, run the following script:
python max_likelihood_estimate.py './results/p3VAE/103'
The figure will be saved in the './results/p3VAE/Figures` folder.
The airborne hyperspectral images acquired during the CAMCATT-AI4GEO experiment in Toulouse, France are publicly available here: https://camcatt.sedoo.fr/
To load and save image patches, use an instance of the GeoDataset class in the data.py file.
Please send any feedback to [email protected]
@article{thoreau2022p,
title={p $\^{} 3$ VAE: a physics-integrated generative model. Application to the pixel-wise classification of airborne hyperspectral images},
author={Thoreau, Romain and Risser, Laurent and Achard, V{\'e}ronique and Berthelot, B{\'e}atrice and Briottet, Xavier},
journal={arXiv preprint arXiv:2210.10418},
year={2022}
}