Table 1.

Results of the deep learning model training on the recognition dataset. SGD, Stochastic Gradient Descent; NAG, Nesterov’s Accelerated Gradient; AdaGrad, Adaptive Gradient algorithm; ResNet, Deep Residual Networks.

Model Type	Model	Training Subset		Validation Subset		Testing Subset
		Loss	Accuracy	Loss	Accuracy	Loss	Accuracy
Pre-Trained Models	AlexNet SGD	0.17	89.35%	0.45	82.87%	0.46	82.73%
	AlexNet NAG	0.19	89.32%	0.47	82.76%	0.47	82.75%
	AlexNet AdaGrad	0.49	88.33%	0.47	82.31%	0.47	82.60%
	GoogLeNet SGD	0.25	90.63%	0.53	83.49%	0.54	83.91%
	GoogLeNet NAG	0.31	92.19%	0.54	83.55%	0.53	83.77%
	GoogLeNet AdaGrad	0.35	90.62%	0.58	83.53%	0.58	83.06%
	ResNet SGD	0.27	84.75%	0.34	85.60%	0.31	84.82%
	ResNet NAG	0.34	84.82%	0.40	85.31%	0.35	85.03%
	ResNet AdaGrad	0.26	85.23%	0.38	84.14%	0.37	83.49%
512 × 512 Models	AlexNet SGD	0.41	89.76%	0.57	81.98%	0.44	84.73%
	AlexNet NAG	0.32	89.89%	0.56	82.03%	0.43	84.03%
	AlexNet AdaGrad	0.51	89.33%	0.60	80.20%	0.46	84.79%
	GoogLeNet SGD	0.42	90.72%	0.79	80.64%	0.60	86.39%
	GoogLeNet NAG	0.35	90.75%	0.78	80.66%	0.58	86.14%
	GoogLeNet AdaGrad	0.48	87.50%	0.76	81.22%	0.48	86.59%
	ResNet SGD	0.62	81.86%	0.36	85.34%	0.29	87.76%
	ResNet NAG	0.45	84.82%	0.29	85.11%	0.26	87.96%
	ResNet AdaGrad	0.50	83.76%	0.32	83.91%	0.33	86.53%
	NutriNet SGD	0.46	88.59%	0.46	80.81%	0.27	86.64%
	NutriNet NAG	0.44	88.53%	0.45	81.06%	0.27	86.54%
	NutriNet AdaGrad	0.44	88.76%	0.46	80.77%	0.26	86.72%
	NutriNet+ SGD	0.41	88.32%	0.45	81.01%	0.27	86.51%
	NutriNet+ NAG	0.45	88.31%	0.45	81.08%	0.27	86.50%
	NutriNet+ AdaGrad	0.42	88.35%	0.45	80.88%	0.28	86.38%