A federated calibration scheme for convolutional neural networks: Models, applications and challenges

Abstract

Deep learning has been created as a practical artificial intelligence strategy that takes various layers of information and gives the best in the effects of different classes. The use of deep learning has indicated exceptional execution in other regions, especially in image clustering, division, and recognition. The ongoing advanced learning strategies implement image clustering, which expects to recognize subsequent-level of classifications. This paper gives a definite audit of different deep arrangements and models featuring attributes of a specific convolutional neural network model. Initially, we depicted the working of Convolutional neural networks and their segments, followed by a point-by-point display of various Convolutional Neural Network models beginning with the old-style LeNet model to AlexNet, GoogleNet, VGGNet, ResNet, DenseNet, Xception, PNAS/ENAS, and EfficientNet. We concluded the significant challenges associated with Spatial Exploitation based Convolutional neural network architecture, Depth Based Convolutional neural network architecture, Multi-Path based Convolutional neural network architectures, and width based Convolutional neural network architectures. A definite summary of the review, including the frameworks, information base, application, and precision for every model, is discussed for serving it as the future scope in the above areas.