Exudate Detection in Fundus Images via Convolutional Neural Network

Exudate detection in fundus images is an important task for the screening of people with diabetic retinopathy. In this paper, Convolutional Neural Network (CNN) is used to detect the exudates in fundus images. An auxiliary loss for classification is designed to better train the CNN architecture. Besides, we use a boosted training method to improve and speed-up the CNN training. The trained model has been evaluated on our own annotated dataset and three public available databases, obtaining an AUC of 0.98, 0.96, 0.94, 0.91 respectively.