Deep Learning in Image Classification: Its Evolution, Methods, Challenges and Architectures

Abstract

AbstractDeep convolutional neural network (CNN) is used in various applications of image processing and computer vision tasks. The ability of deep neural networks to learn automatically from the given data makes it more acceptable. CNN has the power to exploit the spatial and temporal correlation in the data. In this paper, a comprehensive survey gives insight into various application areas which can be further explored. Further, the work focuses on the challenges/issues with deep networks. The analysis and understanding of possible feasible solutions to the aforementioned objective are also given. It focuses on the use and development of CNN architectures and also provides a detailed structural analysis of various CNN architectures. The present work will help the researcher community to analyze and apply suitable deep learning architecture for the target application.KeywordsCNNPre-trained modelsImage classificationData augmentation