By Lynne Schaufenbil
Title: Real-time Analysis of Solar Filaments: the Challenges and the Solutions
Over the past decades, object-detection algorithms have outpaced human capabilities, leading to significant advancements in Computer Vision and Image Processing. However, in scientific imagery, including microscopic, telescopic, and medical images, general-purpose object-detection algorithms fall short. One of the main reasons for this lag is the absence of precise pixel-level object segmentation algorithms, despite their numerous applications in many domains revolving around scientific imagery. One such critical area is Space Weather, where extreme solar events can disrupt vital infrastructures on Earth, like power grids and GPS systems. MLEcoFi, a Machine Learning Ecosystem for Analysis of Filaments, is our data-driven response to this challenge, focusing on solar filaments. Filaments are clouds of ionized gas (plasma) in the solar chromosphere, which are critical for Space Weather Forecasting, as they can tell us about the occurrence of geomagnetic storms.
In this talk, I will share the challenges in developing a proper filament detection and identification system, and how my team has been attacking them. I will review the methods and concepts I have introduced in the past few years towards this goal. These efforts involve the manual annotation of solar filaments captured by the GONG's array of six ground-based observatories, the evaluation of our manual annotation pipeline, devising object-similarity measures sensitive to the fine structures of filaments, the augmentation of filaments, and the evaluation of models' learning process from imbalanced data.
If you are interested in attending, please contact Lynne_Schaufenbil@uml.edu for the Zoom link.