Architecture of Convolutional Neural Networks
Alka Singh, Mayank Deep Khare · Journal of Critical Reviews · 2020
Neural network is a field which comes under a broader field called Machine learning. Neural networks are of different types and these are used for different purposes. Convolutional Neural network is a category of neural networks which is used to play with images. CNN basically deals with image inputs. CNN is buildup of several combinations of convolutional layers and pooling layers. Each of these convolutional layers perform the function of narrowing down the feature groups. When the real time images of images under test are passed on multiple times in a convolutional neural network, then CNN learns a pattern in each image and stores the features which define that object clearly. Real time object detection is successful these days only because of the existence of convolutional neural networks. Connectivity of neurons in a human brain directly relates to the connectivity of layers in a convolutional neural network. CNN is widely used in action recognition, image classification, analysis of a document, human pose estimation and scene labelling.