Comparison of Google Net Convolutional Neural Network to Visual Geometry Group Convolutional Neural Network
Keywords:
Neural Network, Convolution layer, Max pooling, Average pooling Complex cell (C-cell), Simple cell (S-cell), Visual Geometry Group (VGG), Google Net, ImageNet Large Scale Visual Recognition Challenge (ILSVRC)Abstract
Human being has always had the dream of making computer operate with human intelligence in other to solve problem. But it has always been a difficult task to train a system to operate with human intelligence. But AlexNet was able to achieve an impressive result in the ImageNet Large Scale Visual Recognition Challenge using convolutional neural network, which motivated researchers in the development of more improved and robust models and led to the development of Google Net, Visual Geometry Group and other models. In this paper, Google Net which is based on the enhancement of the inception model is compared to Visual Geometry Group which is a very deep convolution in terms of architectural design and performance efficiency. A single inception model of Google Net and a single block of Visual Geometry Group were also implemented to show the differences in their architectural design and mode of operation. The comparison shows that Google Net has a simpler architecture and took lesser time to train. It also outperformed VGG in the 2014 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) with a top-5 error of just 6.67% compared to 7.32% of Visual Geometry Group (VGG).