Deep Learning-Driven Analysis for Ocular Disease Diagnosis

Ms.P.Sujitha, M.Tech, Assistant
Professor, Dept of AI&DS,Annamacharya institute of technology andsciences, Tirupati-517520, AP.
sujithapannapalli@gmail.com
G Hari, UG Scholar,
Dept of AI&DS,Annamacharya institute of technology andsciences, Tirupati-517520, AP.
gowrigariaravind@gmail.com
R Dhanasekhar, UG Scholar, Dept of AI&DS,Annamacharya institute of technology and sciences, Tirupati
517520, AP. dhanasekhar01062005@gmail.com
D Akram, UG Scholar,Dept of AI&DS,
Annamacharya institute of technology andsciences, Tirupati-517520, AP.
akramakram86519@gmail.com
S Mohammad Zunaid, UG Scholar,Dept of AI&DS,
Annamacharya institute of technology and sciences, Tirupati-517520, AP.zunaidshaik48@gmail.com

Abstract— Doctors have tried for a long time to detect eye diseases early enough using fundus images. To be honest, it’s a massive headache tryingto do it manually. It’s slow, it’s costly, and it’s too easy for a human to miss something small. This is why there’s a massive push at the moment for systems that can scan images of the eye and automatically detect diseases immediately. In our research, we’re going deep into the possibilities of a deep learning method for making these diagnoses much more precise. With the massive advances in AI technology that have occurred recently, image classification is now at a stage where it’s actually reliable enough to be used in a real-world setting. We would like to confirm this, and we decided to do so using the ODIR dataset, which contains approximately 6,000 images distributed across eight different categories. We first tried using some classics, such as VGG-19 and ResNet50, but as is the case with this data, there was a small issue. We therefore decided to balance the data so as to create a simple binary classification. We also decided to give the Vision Transformers a try, but we didn't hold back by sticking to the data; we decided to throw in the Local Binary Patterns as well so as to give them a chance to "see" the textures.Keywords— deep learning, transformer networks, ocular diseases, deep learning, medical imaging, computer-aided diagnosis