DRDr: Automatic Masking of Exudates and Microaneurysms Caused by Diabetic Retinopathy Using Mask R-CNN and Transfer Learning

This paper addresses the problem of identifying two main types lesions—Exudates and Microaneurysms—caused by Diabetic Retinopathy (DR) in eyes diabetic patients. We make use Convolutional Neural Networks (CNNs) Transfer Learning to locate generate high-quality segmentation mask for each instance lesion that can be found patients’ fundus images. create our normalized database out e-ophtha EX MA tweak Mask R-CNN detect small lesions. Moreover, we employ data augmentation pre-trained weights ResNet101 compensate dataset. Our model achieves promising test mAP 0.45, altogether showing it aid clinicians ophthalmologist process detecting treating infamous DR.