YOLO-CIANNA: Galaxy detection with deep learning in radio data - I. A new YOLO-inspired source detection method applied to the SKAO SDC1

Home

All issues

Volume 690 (October 2024)

A&A, 690 (2024) A211

Full HTML

Open Access

Table D.1

Detection-only score metrics for various approaches with and without the extra-parameter part of the loss.

Approach	Epoch	AP0.1	AP0.5	AP0.9	$N_{0.1}^{m}$ $N_{0.1}^m$	$N_{0.5}^{m}$ $N_{0.5}^m$	$N_{0.9}^{m}$ $N_{0.9}^m$	Score	AP_{S DC1}	$N_{S D C 1}^{m}$ $N_{S\,DC1}^m$
With param. reference	3000	15.71	13.57	0.70	835789	737056	120467	480439	14.47	679778
With param. no cascade	4200	14.94	13.12	0.69	791421	708597	117005	476943	14.24	676288
With param. patch norm.	3200	15.25	13.05	0.71	811351	711355	117859	461417	14.08	654838
With param. dual norm.	3600	15.96	13.74	0.73	857652	751529	126662	473758	14.35	657853

No param.	2600	15.12	13.05	0.58	801494	709853	108664	N/A	N/A	N/A
No param. more regul.	3000	14.91	13.06	0.73	789935	707086	119189	N/A	N/A	N/A
No param. light	3000	14.93	12.89	0.67	791478	699148	114422	N/A	N/A	N/A
No param. patch norm.	2600	15.62	13.18	0.64	836355	723201	113519	N/A	N/A	N/A
No param. dual norm.	2600	15.84	13.52	0.62	847780	740967	116617	N/A	N/A	N/A

Random 950k	N/A	0.40	0.01	0.00	137877	26416	633	N/A	N/A	N/A
Random 110k	N/A	0.05	0.00	0.00	17152	3306	75	N/A	N/A	N/A

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.