Regular structural analysis of Optic Disc (OD) and Optic Cup (OC) using retinal images can detect irreversible diseases like Glaucoma at an early phase. However, the manual detection of these changes is a tedious task and prone to errors. To alleviate this issue, convolutional neural networks play a vital role in the automated detection of OD and OC. Still, these methods perform a deep-level search which requires a larger dataset and consumes more memory. Therefore, in this article, a robust automated model named Attention Guided Light Weight Res-Net (ALWRes-Net) is implemented to detect patterns of OD and OC from convoluted feature maps. The residual modules of ALWREs-Net effectively restore the shallow features enabling the model to locate the OD and OC pixels more accurately. Apart from this, the attention modules in the skip connection path elevate the discriminative and generalization ability of the model. As a result, at each decoder level, the feature maps are enhanced and focused more on OD and OC regions. The efficiency of the developed ALWRes-Net is computed by using standard datasets such as REFUGE, Rimone-V3, Drishti-GS1, and ORIGA, and attained an average Dice coefficient of 97% for OD and 94% for OC detection.