Image Processing of Fundus Lesions Based on Deep Learning

Researcher, Supervised by Professor Jialin Peng

Content:

• Used public dataset diaretdb as the main data source;

• Recognized four types of diseases in the fundus image, among which each one has some particular areas/features in the image；

• Analyzed the fundus image pathology by using U-Net, U2Net first to test the performance；

• Created U2-Net + PSPNET to test the performance in the datasets；

• Constructed a neural network that could analyze the pathological images of the fundus.