Peningkatan Kontras Citra dengan Metode CLAHE pada Pengklasifikasian Diabetic Retinopathy menggunakan Model InceptionV3

Diabetic retinopathy is a diabetes-related disease that affects the eyes. Early and accurate detection is crucial to improving the chances of recovery. The classification of diabetic retinopathy helps in determining the severity and appropriate treatment. This study aims to enhance image contrast using the Contrast Limited Adaptive Histogram Equalization (CLAHE) method and evaluate its impact on the accuracy of diabetic retinopathy classification using the InceptionV3 model. The dataset used in this study is “APTOS 2019 Blindness Detection,” which consists of 3662 images with 5 classes: No DR, Mild, Moderate, Severe, and Proliferative DR. Three testing scenarios were applied in this study: Scenario 1 without using CLAHE, Scenario 2 using CLAHE first then augmentation, and Scenario 3 augmentation first then using CLAHE. The test results showed that the accuracy obtained was 74% for the scenario without CLAHE, 72% for the scenario with CLAHE before augmentation, and 76% for the scenario with augmentation before CLAHE. These results indicate that using CLAHE after augmentation provides a significant increase in the accuracy of diabetic retinopathy classification.