Fig. 8
From: Automated identification of retinopathy of prematurity by image-based deep learning
Representative images of system-predicted lesion locations compared with those predicted by experts. Box outlines in (a-f) indicate lesion sites. The boxes in red are the lesion locations annotated by retinal experts (gold standard annotations), and the boxes in blue are the lesion locations predicted by the Faster-RCNN. Abbreviation: Faster-RCNN, faster region-based convolutional neural network