An Effective Method for Detecting and Classifying Diabetic Retinopathy Lesions Based on Deep Learning

Table 5

Results obtained by using Kaggle dataset and different pretrained models on the proposed method.

Model	TP	FN	TN	FP	ACC	AUC	SEN
VGG16	971	9	3887	33	99.1	99.9	99.1
VGG19	969	11	3889	31	99.1	99.7	98.9
DenseNet201	919	61	3865	55	97.6	98.6	93.8
DenseNet121	952	28	3894	26	98.9	99.6	97.1
DenseNet169	836	144	3811	109	94.8	98.5	85.3
MOBILENET	684	296	3630	290	88.0	92.6	69.8
NASNet	508	472	3496	424	81.7	83.8	51.8
InceptionV3	674	306	3683	237	88.9	92.3	68.8
InceptionResNetV2	540	440	3726	194	87.0	84.3	55.1
Resnet50	202	778	3142	778	68.2	82.4	20.6