l2 wind direction: README

This repository provides tools for predicting wind direction and estimating uncertainty from SAR images acquired by Sentinel-1 (S1), RADARSAT Constellation Mission (RCM), and RADARSAT-2 (RS2) satellites. It utilizes a trained machine learning model built with PyTorch Lightning and Hydra-Zen

Features

• Prediction Pipeline: Load trained models and make predictions on new data.

• Hydra-Zen Integration: Flexible configuration management.

• Output Options: Generate predictions as a pandas DataFrame or an xarray Dataset.

• Uncertainty Estimation: Predict both wind direction and associated uncertainty.

• Georeferencing Support: Adjust predictions for geographical reference using dataset metadata.

Installation

Clone the repository:

git clone https://github.com/jean2262/l2winddir.git
cd l2winddir

Usage

Command-Line Interface

Run predictions using the predict.py script:

python predict.py --model_path <path_to_model> --data_path <path_to_data> --eval <True/False>

• Arguments:

  • --model_path: Path to the trained model's directory.

  • --data_path: Path to the data file (NetCDF or xarray-compatible format).

  • --eval: If True, outputs a pandas DataFrame; otherwise, modifies and returns the input xarray Dataset.

Example

Command-Line

python predict.py --model_path "/path/to/model" --data_path "/path/to/dataset.nc" --eval True

Programmatic Usage

You can use the make_prediction function directly in Python scripts:

from predict import make_prediction

model_path = "/path/to/model"
data_path = "/path/to/dataset.nc"
result = make_prediction(model_path=model_path, data_path=data_path, eval=True)

print(result)

License

This project is licensed under the MIT License. See the LICENSE file for details.