The Effect of Dimension on Image Recognition using Convolutional Neural Network

Abstract

We study the effect of image dimension on the accuracy and efficiency in image recognition. After preprocessing the dataset., we obtain several datasets with different image dimensions, which can be used to training the model of neural network. By comparing the predicted results and the original labels, as well as recording the accuracy, loss of each epochs, and time-consuming on each training precess, we obtain the scheme for optimal accuracy and efficiency in this task. Our results provide a new perspective on improving the image recognition.