A Brief Journey of Convolutional Neural Networks from 2012 to 2017

With the recent advancements in the field of Deep Learning, largely due to the increase in computational power, the long-standing problem of large-scale image classification has been solved up to a considerable threshold. Since the advent of AlexNet in 2012, there has been an increase in the research and development of various Convolutional Neural Network based model to solve problems in the field of Computer Vision. This may be primarily attributed to the success of AlexNet in successfully classifying 1000 class subset of ImageNet with considerable accuracy. Consequently, many more Convolutional Neural Networks were introduced including VGG Net, Inception Net, ResNet, DenseNet, etc. each with better performance than its predecessor. This paper serves as a brief guide to the architecture of ResNet and some of its predecessors. It will also discuss DenseNet and ResNeXt as improvements of ResNet architecture.