FMARS: Annotating Remote Sensing Images for Disaster Management using Foundation Models

Dataset and code for generating the dataset described in the paper FMARS: Annotating Remote Sensing Images for Disaster Management using Foundation Models.

Note: The dataset is available at Hugging Face Datasets.

Environment Setup

Python Virtual Environment

We recommend using Python version 3.10.12 for this project. Set up a new virtual environment using the following commands:

pyenv install 3.10.12
pyenv global 3.10.12
python -m venv .venv
source .venv/bin/activate

Install the required Python packages with:

pip install -r requirements.txt

Resources

Download necessary metadata using:

wget -q https://github.com/links-ads/igarss-fmars-gen/releases/download/v0.1.0/metadata.zip
unzip metadata.zip
rm metadata.zip

Model Weights

Download the model weights for Grounding DINO and EfficientSAM:

• Grounding DINO:

cd models/GDINO/weights
wget -q https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
cd ../../..

• EfficientSAM: Download the checkpoint from this link and place it in models/efficientSAM/weights/.

Data Acquisition

• Maxar Images:

⚠️ WARNING: Downloading all necessary Maxar images requires more than 900 GB of free space and several hours.

python src/maxarseg/scripts/downloadMaxar.py

• Microsoft Road Detections:

python src/maxarseg/scripts/downloadRoads.py

Usage

You can run the automatic labelling process on a whole event or on a partition of an event using a single command. The event is specified by its index, which can be an integer for a whole event or a decimal for a partition.

python -B -O -m maxarseg --out_dir_root <path/to/output/directory> \
        --config "<path/to/config/file>" \
        --event_ix <event_index>

Parameters

--out_dir_root: Specifies the output directory for the labelling results.

--config: Path to the configuration file (e.g., "./configs/trees_cfg.yaml").

--event_ix: Index of the event to process.

• Use an integer (e.g., 0, 1, 2) to process a whole event.
• Use a decimal from x.0 to x.9 (e.g., 0.0, 0.1, ..., 0.9) to process a partition of an event.

Acknowledgements

FMARS leverages several open-source projects. We extend our gratitude to the authors of the following tools for making their software publicly available:

• Grounding DINO
• EfficientSAM