Glaucoma, a series of progressive eye illnesses, is a primary worldwide health concern. Glaucoma, sometimes known as the "silent thief of sight," progressively affects the optic nerve, resulting in permanent vision loss and, in extreme instances, blindness. It is essential to recognize glaucoma in its earlier stages so that patients can receive treatment sooner and prevent further vision loss. An effective method for detecting glaucoma by analyzing retinal images with the assistance of a deep learning strategy is presented as a potential solution in this article. The framework presented for detecting glaucoma comprises two modules that rely on one another: the Retinal Image Classification Module (RICM) and the Retinal Image Diagnosis Module (RIDM). The retinal image is classified as either a normal or a glaucoma retinal image by the RICM module, which uses the CNN classifier. The RIDM detects the neuro rim region from the glaucoma retinal image by segmenting OD and OC, and the Dual Functional CNN (DFCNN) classifier is proposed to diagnose the severity stages of the glaucoma image based on the feature patterns that are extracted from the neuroretinal rim in the glaucoma image. Both low- and high-resolution retinal image datasets, known as HRF and PAPILA, are utilized in this study to investigate the proposed approaches for glaucoma identification and severity estimate. Compared to other methods considered to be state-of-the-art, the simulation's findings show that it is successful. Ophthalmologists benefit from the suggested model since it assists them in effectively recognizing glaucoma in patients, which in turn allows for improved diagnosis and the prevention of premature vision loss.