• Vaibhav Kumar Dubey
• Punit Kumar Johari

Deep Convolutional Neural Networks (CNNs) are specialized Neural Networks capable of proving state-of-the-art results based on various benchmarks. CNN has somewhat powerful learning ability which is achieved by using various feature extraction levels that can learn the structural or hierarchical representations from the data. With the exponential increase of large amount of data and certain improvements in the computation power of hardware’s there has been an increment in the research of CNN which has resulted in development of various architectures of CNN. Recent trend in developing deep CNN architectures has improved the performance on various computer vision related tasks based on innovations in ideas behind architectures and parameter optimization. Due to these various architectural designs has been explored based on different activation and loss functions, regularization, parameter optimization and restructuring the processing units. The latter has proved to be the major improvement in the representational capacity of CNN along with using block as a structural unit instead of a layer. This paper thus focuses on different versions of CNN architectures reported recently, covers the understanding of elementary concepts of CNN Components and throw some lights on application and challenged in CNNs.

Deep Learning, Convolutional Neural Networks, VGG, AlexNet, ResNet and DenseNet.

"Convolutional Neural Networks and Architectures: A Brief Overview", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 5, page no.341-350, May-2019, Available :http://www.jetir.org/papers/JETIR1905652.pdf

"Convolutional Neural Networks and Architectures: A Brief Overview", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 5, page no. pp341-350, May-2019, Available at : http://www.jetir.org/papers/JETIR1905652.pdf