Table 9.

Specifications of the utilized FT CNN models.

CNN Model	Disk Size (MB)	Layers	Parameters (Total)	Parameters (Trained)	Parameters (Non-Trained)	Reduced (%) in Training Parameters
ResNet50	96	50	25,600,000	4096	25,595,904	99.98
AlexNet	227	8	61,000,000	8192	60,991,808	99.99
InceptionV3	89	48	23,900,000	4096	23,895,904	99.98
ResNet101	167	101	44,600,000	4096	44,595,904	99.99
GoogleNet	27	22	7,000,000	2048	6,997,952	99.97
VGG16	515	16	138,000,000	8192	137,991,808	99.99
DarkNet53	155	53	41,600,000	2048	41,597,952	99.99
Xception	85	71	22,900,000	4096	22,895,904	99.98
InceptionResNetV2	209	164	55,900.00	3072	55,896,928	99.99
MobileNetV2	13	53	3,500,000	2560	3,497,440	99.93
NasNetMobile	20	*	5,300,000	2112	5,297,888	99.96
DarkNet19	78	19	20,800,000	2048	20,797,952	99.99
ResNet18	44	18	11,700,000	1024	11,698,976	99.99
DenseNet201	77	201	20,000,000	3840	19,996,160	99.98
NasNetLarge	332	*	88,900,000	8064	88,891,936	99.99
Places365-GoogleNet	27	22	61,000,000	2048	60,997,952	99.99
ShuffleNet	5.4	50	7,000,000	1024	6,998,976	99.99
SqueezeNet	5.2	18	1,240,000	1024	1,238,976	99.92
VGG19	535	19	144,000,000	8192	143,991,808	99.99

* The NasNetLarge and NasNetMobile models do not comprise a linear structure of CNN modules.