Automatic Detection of AMD and DME Retinal Pathologies Using Deep Learning

Table 4

Obtained results on the Duke dataset.

Model	Accuracy	Precision	Recall	score
Fine-tuning of Xception	96.83%	96.66	97%	97%
Feature extraction from “sepconv2_bn”+CNN	98.02%	98.33	98%	97.66%
BCNN (Xception, Xception)	97.84%	97%	97%	97%
Fine-tuning of Inception-ResnetV2	93.43%	93.66%	93.33%	93.33%
Features extraction from “block8_1_ac”+CNN	97.70%	98%	97.66%	97.66%
BCNN (Inception-ResnetV2, Inception-Resnet-V2)	95.55%	95%	95%	95.66%
The from scratch “OCTorch-Net”	99.68%	99%	96%	97%