Develops and deploys deep learning models for solar physics using preprocessed Sun and Space Weather data.
SSW-ML is a specialized toolkit for solar physicists and data scientists working with solar observation data. It streamlines the creation of machine learning pipelines by providing domain-specific PyTorch and TensorFlow dataloaders for FITS files, standardizing preprocessing steps like log-scaling and normalization, and offering reference architectures for tasks such as instrument-to-instrument translation, flare prediction, and coronal hole segmentation. It bridges the gap between raw solar data and modern deep learning frameworks, enabling faster iteration on complex heliophysics research models.
주요 기능
01Comprehensive evaluation metrics including SSIM and PSNR for solar image translation
02Domain-specific normalization and log-scaling for high-dynamic-range solar imagery
03Custom PyTorch and TensorFlow DataLoaders specifically designed for FITS file formats
04Integrated workflows for multi-wavelength observation pairing and training
05Pre-built architectures for U-Net, ResNet, and CNN+LSTM solar models
060 GitHub stars
사용 사례
01Predicting solar flare events using EUV time-series and image data
02Automated segmentation and detection of coronal holes and active regions
03Translating solar observations between different instruments like STEREO and SDO
