Glaucoma, an optic nerve disease resulting in blindness if left untreated, is a difficult condition in healthcare in view of its diagnostic difficulties. Past approaches are based on assessment of the fundus images and the size of the cup and the disc, thickness of the rim and other abnormalities present in the eyes. Multifaceted developing AI prospects have potential to improve glaucoma identification. This research aims at implementing the best of feature learning on UNet + + and Capsule Network (CapsNet) for better diagnostic results. For semantic segmentation, UNet + + is used to accurately outline ODs and OCs-significant features in glaucoma diagnosis. CapsNet further performs this by capturing hierarchical structures and proves to be more sensitive towards the glaucomatous changes than the conventional Convolutional Neural Networks. To get improved image quality and features, a standard pre-processing method, such as Histogram Equalization and Contrast Limited Adaptive Histogram Equalization (CLAHE), are used in this paper to preprocess retinal images. The proposed hybrid model is then trained and tested on ten benchmark sets and achieves high accuracy in optic disc and cup segmentation and better performance in glaucoma detection than other presented approaches. Performance evaluation suggests good diagnostic ability which opens up the possibility of an automated system that helps clinicians diagnose early glaucoma. This application of UNet + + and CapsNet shows the current possibility of glaucoma diagnosis using a new and more efficient method and a relatively small but powerful potential to prevent blindness by diagnosing glaucoma at an early stage. However, the study notes that the application of AI has revolutionized ophthalmic health care.