Detecting eye diseases like Glaucoma and Diabetic Retinopathy early is crucial to prevent irreversible vision loss; these conditions, characterized by optic nerve damage are recognized as leading causes of blindness. Fundus photography involves capturing images of the back of the eye and it plays a crucial role in diagnosing and monitoring various eye conditions, aiding in the early detection and management of these diseases and their subtypes, such as the different stages of glaucoma and diabetic retinopathy. Existing methods for analysing fundus images rely on traditional image processing techniques like filtering and transformation, and classification methods like CNN or SVMs. These methods do not reliably capture data, leading to lower accuracy in identification. In this study, we utilized fundus images from the FundusImage1000 dataset obtained from Kaggle, which contains a comprehensive collection of images capturing various stages of glaucoma and diabetic retinopathy. This study proposes an ensemble method for glaucoma and diabetic retinopathy detection using fundus images. By utilizing Grey-Level Co-occurrence Matrix for texture analysis, this method integrates classification models XGBoost and Light GBM to make accurate predictions for early identification of the subtypes of these two diseases, including the various stages such as Mild, Moderate, Advanced, Proliferative or Nonproliferative. This combination of texture analysis and ensemble learning yields effective results in automated disease identification. The proposed approach also achieves a high accuracy of 99%, offering a valuable tool for medical professionals to improve diagnostic outcomes and patient care.