Skip to main content
. 2022 Nov 8;9:1050436. doi: 10.3389/fmed.2022.1050436

Figure 1.

Figure 1

The workflow of DR-detector framework for detection of DR using a combination of TL and lesion features. (A) Extracting TL features from VGG-16 and ResNet-50 models. (B) Extracting MAs, EXs, and HEMs lesion features from retinal fundus images using U-Net semantic segmentation models. (C) Flat combination of TL and lesion features in DR-detector framework.