Segmentation is one of the most significant processes in identifying the diseases. Glaucoma disease is detected by segmenting Optic Disc (OPdisc) and Optic Cup (OPcup) from a fundus image. In disease detection method, accuracy of segmentation plays a vital role. Segmentation process is more time consuming task because of large dataset. For avoiding that, an automatic segmentation tool is needed. In this paper, the automatic segmentation is proposed through a Deep Learning based CNN model. The OPdisc and OPcup are segmented by using a hybrid Channel Attention Gate- Squeeze Excitation Parallel Pooling Statistical Map (CAG-SEPPSM) embedded using Convolutional Neural Networks (CNN). In segmentation of OPdisc, a new Attention Gate and in segmentation of OPcup, Squeeze-Excitation Parallel Pooling Statistical Map block was developed. The dataset which are used for testing the proposed method are DRISHTI-GS database and RIM-ONE v.3 database. The proposed segmentation method outperform when compared to the existing methods in terms of Dice Coefficient (DC), Intersection Over Union (IOU) and Accuracy.