A comparison between CPU and GPU for image classification using Convolutional Neural Networks
Anusha Jayasimhan, P. Pabitha · 2022 International Conference on Communication, Computing and Internet of Things (IC3IoT) · 2022
Deep learning has emerged as an effective solution for analyzing complex datasets such as images, videos, text and speech. Today, one of the most popular algorithms to perform image classification is Convolutional Neural Networks(CNN) and its various implementations such as VGG, ResNet, Inception etc. The recent advances in hardware has led to emergence of Graphical Processing Units (GPU) as a solution for speeding up the process of executing complex deep learning algorithms. Although GPU offers massive parallelism through a large number of cores, it is not always the optimum choice for executing training of all deep learning models. This work performs a comparative analysis on CPU and GPU for two datasets using two different Convolutional Neural Network models. The experiments have been conducted on three different types of datasets and each trained with a separate CNN model. All three experiments indicate that the CPU trains upto 1.5x-3x times faster than the GPU.