AL4EO

AL4EO/qgis-plugin is a QGIS plug-in to run Active Learning techniques on Earth observation data.

The goal of Active Learning is to interactively and iteratively build optimal training data sets for supervised learning. For further details on Active Learning for remote sensing data, see our comparative review here.

If you use AL4EO as part of a published research project, please see the Reference section below to cite our paper.

To run benchmark experiments, see the benchmark branch of this repository.

Despite our constant efforts to improve the user experience, AL4EO can still be tricky to use. Feel free to open an issue on Github or directly send us an email at [email protected].

Requirements

QGIS

To install the plug-in, compress the qgis_plugin folder into a zip file and use the QGIS plug-in manager.

To plot pixels spectra, we strongly recommend to install the QGIS Temporal/Spectral Profile Tool.

Python dependencies

AL4EO is compatible with Python >= 3.7 and <=3.10 and PyTorch 1.10.0+.

The easiest way to install the backend code is to create a AL4EO conda virtual environment and to install dependencies using:

pip install -r requirements.txt

(on Windows you should use pip install -r requirements.txt -f https://download.pytorch.org/whl/torch_stable.html)

Usage

Active Learning Framework

The two major building blocks of the Active Learning Framework are:

• the acquisition function, that tells how informative the unlabeled pixels are,
• the oracle, that labels the queried pixels (the pixels that maximise the acquisition function).

Active Learning techniques iteratively query pixels as follows:

At this time, the qgis plug-in handles three acquisition functions:

• Breaking Ties, an inter-class uncertainty heuristic,
• BALD, an epistemic uncertainty heuristic,
• Core-Set (only the k-center greedy algorithm), a representativeness heuristic.

See more acquisition functions in AL4EO/benchmark.

Models

Acquisition functions are often based on machine learning classifiers that are defined in AL4EO/learning/models.py.

Data

AL4EO plug-in handles data in the ENVI format. Your data folder should be organized as follows:

Data
├── image.tiff
├── image.hdr
├── ground_truth.shp

The image.tiff file is the image to annotate. Its image.hdr header should contain a bbl attribute that lists the bad band multiplier values of each band in the image, typically 0 for bad bands and 1 for good bands. The ground_truth.shp file is the initial ground truth, saved as a shapefile with a "Material" field.

Examples of files are provided in the Demo folder.

How to start

Launch the backend in your AL4EO conda virtual environment with the following command line:
python -m server

In QGIS, the Q button opens a window where you can select:

• the layers on which you run the query,
• the acquisition function,
• various hyperparameters,
• preprocessing options (as introduced in here).

Once the query is completed (see progress in the terminal), a history layer pops up that contains the selected pixels, shown as red points.

Reference

This toolbox was used for our review paper in IEEE Geoscience and Remote Sensing Magazine:

R. Thoreau, V. Achard, L. Risser, B. Berthelot and X. Briottet, "Active Learning for Hyperspectral Image Classification: A Comparative Review," in IEEE Geoscience and Remote Sensing Magazine, doi: 10.1109/MGRS.2022.3169947.

@ARTICLE{9774342,  
author={Thoreau, Romain and Achard, Veronique and Risser, Laurent and Berthelot, Beatrice and Briottet, Xavier},  
journal={IEEE Geoscience and Remote Sensing Magazine},   
title={Active Learning for Hyperspectral Image Classification: A Comparative Review},   
year={2022},    
pages={2-24},  
doi={10.1109/MGRS.2022.3169947}}

If you use the preprocessing method implemented in AL4EO as part of your work, please cite the following paper:

Thoreau, R., Achard, V., Risser, L., Berthelot, B., and Briottet, X.: ACTIVE LEARNING ON LARGE HYPERSPECTRAL DATASETS: A PREPROCESSING METHOD, Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B3-2022, 435–442, https://doi.org/10.5194/isprs-archives-XLIII-B3-2022-435-2022, 2022

@Article{isprs-archives-XLIII-B3-2022-435-2022,
AUTHOR = {Thoreau, R. and Achard, V. and Risser, L. and Berthelot, B. and Briottet, X.},
TITLE = {ACTIVE LEARNING ON LARGE HYPERSPECTRAL DATASETS: A PREPROCESSING METHOD},
JOURNAL = {The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences},
VOLUME = {XLIII-B3-2022},
YEAR = {2022},
PAGES = {435--442},
URL = {https://www.int-arch-photogramm-remote-sens-spatial-inf-sci.net/XLIII-B3-2022/435/2022/},
DOI = {10.5194/isprs-archives-XLIII-B3-2022-435-2022}
}