Speaker
Description
Astronomical facilities generate ever-increasing data volumes, rapidly approaching the exascale. This is especially true for modern radio-interferometers for which the foreseen data rate raises strong concerns regarding the scaling capability of classical analysis methods.
In this talk, I will introduce YOLO-CIANNA (Cornu et al. 2024), a deep-learning object detector designed for astronomical images, and present how we used it to detect and characterize galaxies in simulated 2D continuum images and HI emission cubes from the first two editions of the SKAO Science Data Challenges (Bonaldi et al. 2020, Hartley et al. 2023). Thanks to this approach, we improved the SDC1-winning score by 139% and enabled team MINERVA to achieve first place in the SDC2. I will then present preliminary results from the application of our method to continuum images from LOFAR (LoTSS DR2) and HI cubes from MeerKAT (preliminary LADUMA), and discuss how we plan to generalize its application to other surveys.
