Pathological Myopia Image Recognition Strategy Based on Data Augmentation and Model Fusion

Table 2

VGG-16 training results on 13 datasets.

No.	Dataset	Accuracy	Loss
1	PALM-Training1600-overturning-dimming-imgaug2	0.95858336	0.18674079
2	PALM-Training3200-overturning-noise-color-cropping-deforming-dimming	0.95550001	0.27185006
3	PALM-Training1600-overturning-cropping-deforming	0.95266668	0.16523545
4	PALM-Training800-color	0.95033336	0.17019135
5	PALM-Training800-dimming	0.94875002	0.17919912
6	PALM-Training800-cropping	0.94625	0.18303553
7	PALM-Training1600-overturning-dimming	0.94525003	0.23124305
8	PALM-Training1600-overturning-noise-color	0.94008333	0.21350351
9	PALM-Training800-overturning	0.93858335	0.20894363
10	PALM-Training800-deforming	0.93708334	0.20814224
11	PALM-Training800-noise	0.93608335	0.26124661
12	PALM-Training800-imgaug1	0.93391667	0.19876853
13	PALM-Training400	0.93016667	0.19310093