Table 3. Overall training and validation accuracies for different CNN architectures with variable numbers of layers.

Architecture	# of Layers	Train Acc.	Validation Acc.
ResNet	18	99.10%	92.02%
	34	99.01%	93.16%
	50	98.43%	90.87%
	101	99.46%	95.44%
	152	99.31%	93.54%
Wide ResNet	50	99.28%	93.92%
	101	99.48%	93.92%
ResNext	50	98.77%	94.30%
	101	99.37%	95.82%
DenseNet	121	99.09%	92.02%
	161	99.50%	94.30%
	169	99.50%	95.06%
	201	99.88%	96.20%