Glaucoma is an eye condition frequently associated with diabetes. It is a major cause of irreversible blindness, particularly in Africa, where medical resources are often limited. Senegal, like many other African countries, faces specific challenges in detecting and treating glaucoma due to these constraints. Early detection of glaucoma is crucial for preventing vision loss, but it is complicated by the absence of symptoms in the early stages of the disease. In this paper, we evaluates the effectiveness of various machine learning models for the early detection of glaucoma from retinal images. We used transfer learning with the pre-trained VGG16 model to extract features from the images, followed by Principal Component A nalysis (PCA) to reduce dimensions while preserving essential information. Three classical machine learning models were then trained on these features: Support Vector Machines (SVM), Decision Trees, and K-Nearest Neighbors (KNN). Performances of these models was compared, and the results show that the SVM offers the best performance among those evaluated on the test set. Bases on SVM model, we used GridSearchCV to identify and optimize the hyperparameters to enhance the performance of the SVM model. Machine learning models have significant potential to improve glaucoma screening, making diagnoses more accessible and reliable, especially in contexts like Senegal, where medical resources are limited.