Table 1.

Autoencoder model comparison with the previous works.

Authors	Framework	Data	Comparison
Ji et al. [15]	MLP	MNIST $\left(28, 28, 1\right)$ CIFAR10 $\left(32, 32, 3\right)$	Small dimensions High-contrast Simple model design
Betechuoh et al. [16]	MLP	HIV records	Estimation Function similar to the Equation (4)
Wang et al. [17]	CNN	BSDS500 $\left(481, 321, 1\right)$	Small dimensions High-contrast
Pihlgren et al. [18]	CNN	STL-10 $\left(96, 96, 3\right)$ SVHN $\left(32, 32, 3\right)$	Small dimensions High-contrast Simple model design
Dumas et al. [19]	CNN	BSDS300 $\left(481, 321, 1\right)$	Small dimensions High-contrast Image normalization preprocessing
Khan et al. [20]	CNN	Applied CICFlowMeter-V3 model to extract the cyberattack features from the CSE-CIC-IDS2018 dataset	Small dimensions Not an image source No image preprocessing required
Ramalho et al. [21]	CNN	The authors construct virtual scenes with multiple 3D views	Not an image source No image preprocessing required Generating synthetic 3D images