Diabetic Retinopathy Diagnosis with InceptionResNetV2, Xception, and EfficientNetB3 Mukkesh Ganesh, Sanjana Dulam, and Pattabiraman Venkatasubbu 1 Introduction Diabetic retinopathy (DR) is a diabetic condition that affects the eyes. Worldwide one-third of the estimated diabetic population show signs of DR. Elevated sugar levels in the blood can lead to the blockage of blood vessels in the retina. This condition is termed as non-proliferative DR (NPDR) which could worsen to pro- liferative DR (PDR). If left untreated, scar tissues stimulated by the growth of new blood vessels may cause the retina to detach from the back of your eye which can cause complete blindness. Hence, early diagnosis is critical to the mitigation of this medical condition. Over the past decade, deep learning (DL)-assisted diagnostic systems have risen in number and have outperformed the traditional image processing-based systems. From detecting cancerous tumors in lungs and breast scans to the diagnosis of COVID-19 from CT scans, this technology has gained wide acceptance within the medical field. Rapid innovation in the deep learning-based computer vision has given rise to powerful neural network archi- tectures which have greatly enhanced the performance of these models. In much more recent years, researchers have started to use the power of transfer learning to make models converge faster and better for tasks that previously had limited training resources. In this paper, we will be utilizing transfer learning to explore and compare the performance of state-of-the-art neural network architectures for diagnosing the severity of DR from retinal fundus images. For training and validating these M. Ganesh (&) Á S. Dulam Á P. Venkatasubbu Vellore Institute of Technology, Chennai, Tamil Nadu, India e-mail: g.mukkesh2017@vitstudent.ac.in S. Dulam e-mail: sanjana.dulam2017@vitstudent.ac.in P. Venkatasubbu e-mail: pattabiraman.v@vit.ac.in © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 R. R. Raje et al. (eds.), Artificial Intelligence and Technologies, Lecture Notes in Electrical Engineering 806, https://doi.org/10.1007/978-981-16-6448-9_41 405