Glaucoma ranks among the primary causes of blindness worldwide. The evaluation of the optical cup-to-disc ratio (CDR) using ophthalmoscopy is the popular diagnostic technique for detecting glaucoma. The CDR is often assessed manually by ophthalmologists, leading to possible inconsistencies due to differences in skill, experience, and interpretation of precise image details. In this work, we introduce an innovative method for addressing glaucoma detection that focuses on extracting relevant image features from a specific Region of Interest (ROI) surrounding the Optic Disc (OD). The OD region in the fundus images is identified using a mask generated by a pre-trained U-Net architecture. To enhance image features in this area, the extracted ROI images are preprocessed using Contrast Limited Adaptive Histogram Equalization (CLAHE) technique. The preprocessed images are further used to train an Instance Aware Attention Vision Transformer (IAA-ViT) architecture for glaucoma screening. The IAA-ViT architecture employs a self-attention mechanism that extracts features from individual image patches to categorize different stages of glaucoma. We evaluated the efficacy of the proposed model in three publicly available datasets, namely ACRIMA, Drishti-GS, and RIMONE. We also addressed the class imbalance in the training datasets by integrating the polyFocalLoss function into the proposed pipeline. Our model has shown comparable performance with other state-of-the-art techniques in identifying glaucomatous eyes from normal, healthy images.