prof. Annie Zavagno (Laboratoire d’Astrophysique de Marseille, France)

Deep learning for detecting galactic interstellar filaments

Filaments are sites of star formation in galaxies. The study of their properties has attracted much attention since the Herschel far infrared telescope revealed their ubiquity in the interstellar medium of our galaxy. Moreover, the formation of stars can be studied through the properties of filaments that host this process. However, in order to study these properties, the filaments must first be extracted from the data.
Filaments are complex structures that form and evolve within the interstellar medium. Physical conditions (local density, temperature and magnetic field intensity) strongly influence their properties. Unbiased detection is essential to understanding their formation and evolution and tracing their life cycle, which includes the star formation process.
Classical detection algorithms suffer from various biases that hinder unbiased detection of filaments. In this talk, I will present the results of our work exploring the potential of machine learning to detect such filaments. I will also discuss the limitations of this work and present some of the approaches we are considering.