سعيد سهرابي امين

Diabetic Retinopathy (DR) is a common complication of diabetes that can lead to blindness, making its early detection highly significant. This research focuses on developing an automated system for classifying five severity levels of diabetic retinopathy using deep learning techniques. The dataset utilized consists of 3,662 retinal images from the Kaggle platform, which exhibits an imbalanced class distribution. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) was employed, an‎d the DenseNet-121 architecture with pre-trained ImageNet weights was selec‎ted as the base model. The model was trained using the Adam optimizer, the Sparse Categorical Crossentropy loss function, an‎d advanced configurations such as EarlyStopping an‎d ModelCheckpoint. eva‎luation results demonstrated that the proposed model achieved an accuracy of 97%, confirming its effectiveness as a method for early detection of diabetic retinopathy. This study also includes error analysis, comparisons with similar methods, an‎d suggestions for future research.

يادگيري عميق - طبقه بندي بيماري- رتينوپاتي ديابت

Disease Classification- Deep Learning- Diabetic Retinopathy

Saeed Sohrabi Amin

Saeed Sohrabi Amin