Performance Comparison of Deep Neural Networks on Image Datasets

With the introduction of convolutional neural networks, error rate of 2.5 percent has been successfully achieved on the image classification. In this paper, four popular convolutional neural network based models namely, VGG16, mobileNet, Resnet50 and InceptionV3 have been considered for the comparative study. The efficiency of the considered models is evaluated on various image classification datasets namely; cats and dogs for binary classification and plant seedling dataset for multiclass classification in terms of accuracy.