<p>Segmentation of Optic Cup (OC) and Optic Disc (OD) from retinal images is a primary step in the analysis of retinal disorders like Glaucoma. Precise extraction of these structures is difficult, due to their complex and irregular shapes. With the recent advancement of Artificial Intelligence (AI) in the healthcare system, a robust deep learning model termed modified Multi-view Attention and Fusion based U-Net (MAF-U-Net) is proposed for the accurate detection of the OD and OC pixels from input retinal images. The main novelty of the developed model is integrated with a multi-view attention module (MAM) and partial parallel decoder module (PPD) to obtain multiscale feature maps from different views for accurate identification of the recursive patterns of OD and OC from fundus images. Furthermore, fusion module (FM), and fusion convolution pyramid pooling (FCPP) modules jointly extract and enhance the most relevant information for the detection of OD and OC regions. Finally, the binary masks of OD and OC are generated by applying a threshold of 0.95 on the probabilistic maps obtained from MAF-U-Net. The competence of the modified MAF-U-Net is assessed using various standard datasets like ORIGA, RIM-ONE, REFUGE &amp; DRISHTI-GS1 to validate the robustness of the developed MAF-U-Net on variable input images. It attained an average dice coefficient of 0.99 ± 0.05 and an accuracy of 0.99 ± 0.03 in locating the pixels of both OD and OC from input images proving the reliability of the network. The developed MAF-U-Net model is effective and can detect the recurrence patterns of OD &amp; OC on retinal images allowing the expert doctors to analyze the Glaucoma more precisely.</p>