This dataset represents the largest collection of manually annotated filaments from H-Alpha observations captured by the Global Oscillation Network Group (GONG). It contains 10,244 annotated filaments from 1,593 observations spanning the years 2011 through 2022. This is the result of 1,066 person-hours of annotation.

This ML-ready dataset is publicly accessible for the community and is ready to be used for training complex machine learning models to enhance our understanding of solar filaments.

We developed a domain-independent Python package to facilitate the preprocessing routines required in preparation of any multi-class, multivariate time series data. It provides a comprehensive set of 48 statistical features for extracting the important characteristics of time series. The feature extraction process is automated in a sequential and parallel fashion, and is supplemented with an extensive summary report about the data. Using other modules, different data normalization methods and imputations are at users' disposal. To cater the class-imbalance issue, that is often intrinsic to real-world datasets, a set of generic but user-friendly, sampling methods are also developed.

We provide a large image parameter data set extracted from the Solar Dynamics Observatory (SDO) mission's Atmospheric Imaging Assembly (AIA) instrument, for the period of 2011 January through the current date, with the cadence of 6 minutes, for nine wavelength channels. The volume of the data set for each year is just short of 1 TiB. Toward achieving better results in the region classification of active regions and coronal holes, we improve on the performance of a set of 10 image parameters, through an in-depth evaluation of various assumptions that are necessary for calculation of these image parameters. Then, where possible, a method for finding an appropriate setting for the parameter calculations was devised, as well as a validation task to show our improved results. In addition, we include comparisons of JP2 and FITS image formats using supervised classification models, by tuning the parameters specific to the format of the images from which they are extracted and specific to each wavelength. The results of these comparisons show that utilizing JP2 images, which are significantly smaller files, is not detrimental to the region classification task that these parameters were originally intended for. Finally, we compute the tuned parameters on the AIA images and provide a public API to access the data set. This data set can be used in a range of studies on AIA images, such as content-based image retrieval or tracking of solar events, where dimensionality reduction on the images is necessary for feasibility of the tasks.