Deep Learning technique has emerged as a successful approach in applications with complex functions. It is used in the areas of computer vision, speech recognition and image processing applications. The most popular deep learning architecture is Convolution Neural Network (CNN). The proposed method a pre-trained CNN based on GoogLeNet and VGGNet is used to classify the retinal diseases such as Age-Related Macular Degeneration (AMD) and Diabetic Macular Edema (DME). Spectral Domain Optical Coherence Tomography (SD-OCT) images are used for identifying the retinal pathologies. Pre-trained GoogLeNet and VGGNet predict the nature of all the subjects of the test images. 94% of accuracy, 95% specificity and 96% sensitivity for the three classes are achieved using GoogLeNet and 96% of accuracy, 99% specificity and 96% of sensitivity for the three classes are achieved using VGGNet.