YOLO-CIANNA-3D: Galaxy detection with deep learning. SKAO SDC2 related codes, models and predicted source catalogs.

This deposit contains various files associated with the article Cornu et al. (2026, A&A) entitled “YOLO-CIANNA: Galaxy detection with deep learning in radio data. II. Winning the SKA SDC2 using a generalized 3D-YOLO network”. The aim of this deposit is to publish, reference, and archive codes, trained models, ancillary data, and source catalogs associated with the article to ensure reproducibility and enable future works. The archive is organized into sub-folders as follows: codes: Source codes and script files allowing to reproduce the results presented in the article (mostly Python scripts and notebooks). models: Trained models save states (final weights) in binary format. metadata: Subsets of quantities that are required for using each trained model (mostly normalization scaling factors). catalogs: All source catalogs presented in the article in a plain text format, obtained from using the provided models. All file names for models, catalogs, and metadata should be self-explanatory and correspond to the naming convention from Cornu et al. (2026).See the README.md file for more details on how to use the content of this deposit.