Table 5 Detailed Retrieval Results for Normal Queries: Retrieval mAPs are evaluated and compared against four baseline models under different conditions for the A-MRI dataset.

Backbone Network	Adversarial Training	Top k	Hashcode Length
Backbone Network	Adversarial Training	Top k	8 bits	16 bits	32 bits	64 bits
AlexNet	Before	100	0.79	0.87	0.84	0.84
		200	0.78	0.85	0.83	0.75
		500	0.74	0.76	0.73	0.75
		1000	0.70	0.70	0.66	0.65
	After	100	0.69	0.80	0.78	0.66
		200	0.66	0.75	0.73	0.58
		500	0.66	0.70	0.66	0.56
		1000	0.64	0.65	0.65	0.54
VGG 16	Before	100	0.77	0.88	0.86	0.77
		200	0.77	0.84	0.75	0.74
		500	0.77	0.78	0.75	0.74
		1000	0.68	0.74	0.73	0.67
	After	100	0.70	0.81	0.84	0.75
		200	0.68	0.77	0.72	0.71
		500	0.67	0.69	0.63	0.58
		1000	0.64	0.67	0.59	0.58
DenseNet 121	Before	100	0.78	0.81	0.81	0.76
		200	0.77	0.80	0.79	0.66
		500	0.75	0.77	0.74	0.64
		1000	0.69	0.71	0.67	0.62
	After	100	0.78	0.84	0.79	0.71
		200	0.74	0.83	0.78	0.65
		500	0.73	0.80	0.70	0.63
		1000	0.63	0.77	0.66	0.60
DensNet 201	Before	100	0.79	0.86	0.82	0.71
		200	0.78	0.83	0.81	0.69
		500	0.77	0.80	0.77	0.68
		1000	0.77	0.80	0.76	0.67
	After	100	0.79	0.86	0.80	0.69
		200	0.78	0.84	0.79	0.67
		500	0.76	0.81	0.78	0.65
		1000	0.76	0.74	0.76	0.65
ConvNeXt	Before	100	0.81	0.91	0.87	0.85
		200	0.80	0.89	0.85	0.73
		500	0.78	0.88	0.83	0.72
		1000	0.72	0.79	0.80	0.71
	After	100	0.82	0.88	0.84	0.81
		200	0.80	0.85	0.82	0.71
		500	0.76	0.80	0.78	0.68
		1000	0.71	0.72	0.69	0.65

Quick links

Search