RADCAT (Radio Analysis of Duplicates CATalogue)

RADCAT is an amulgimation of existing catalogues in literature where the usage of the Vertex Cover Heuristc is employed to remove duplicate sources during catalogue compilation. Furthermore, RADCAT is comprised out of radio sources that have been observed not only in the FIRST and NVSS surveys, but also the LoTSS survey. RADCAT was used to construct six datasets, and as such its purpose is to facilitate various machine learning models that can leverage multi-modal survey inputs to enhance classification accuracy. This includes testing different transfer learning approaches across surveys with varying resolution and sensitivity, as well as exploring cross-survey image generation techniques. The six datasets that depict the sources catalogued in RADCAT are: RADCAT-F (Image dataset with fixed pixel resolution across surveys). RADCAT-Fc (RADCAT-F with sigma clipping applied). RADCAT-V (Image dataset with varying pixel resolution across surveys). RADCAT-Vc (RADCAT-V with sigma clipping applied). RADVIS (Reverse engineered visibilities dataset). RADVISc (RADVIS with sigma clipping applied to the images before reverse engineering visibilities). Column Headings of RADCAT.csv: Index - The source identifier associated with the fits filename. RAh - Hour part of the right ascension. RAm - Minute part of the right ascension. RAs - Second part of the right ascension. DECsign - Positive or negative sign for declination. DECd - Magnitude of degree part of declination. DECm - Arcminute part of declination. DECs - Arcsecond part of declination. Catalog - Radio source origin. Type - RADCAT morphological label. RA/deg - Right ascension in degrees. DEC/deg - Declination in degrees. Divergence - Morphological labels of source as per the constituent catalogues of RADCAT. Original - Morphological label of source as per the source's original catalogue. Supplemental files are available on the GitHub repository. data_excluded.zip - All sources excluded during the construction of the final RADCAT, due to reasons titled in the filenames. graph_theory_approach.py - The noval approach to radio astronomy at removing duplicate sources. read_augmentations.ipynb - An example of how to read the augmentations files inside the dataset zip files an how to use them. source_ids_test.txt - The source ids that correspond to a RADCAT row entry that were used in the test set during research. source_ids_val_train.txt - The source ids that correspond to a RADCAT row entry that were used in the training set during research. Catalogue References.txt - References to each catalogue used to create RADCAT. threshold_search.csv - The source pairs and their distances that fell between the interval [0.005, 0.072] degrees. The 'same' or 'diff' tag was manually assigned upon visual inspection. This aided in obtaining a threshold of 0.03 degrees.