Abstract: In this computing era, researchers are aiming to mimic the human brain and one such field of research is Deep learning (DL). Computer vision which is an application of deep learning uses computer to derive meaning full information from images. These types of computer vision tasks are achieved by Convolutional neural network (CNN), a branch of DL that perform mathematical calculation on images for classification. Object detection and image recognition have been the main motivation for CNN. Many researches have been carried over CNN to create an architecture with less computational cost and high performance. This paper attempts to give a comprehensive survey of the main three architectures namely, GoogleNet, AlexNet, and LeNet. A detailed explanation of the various architectures is discussed.