Performance Analysis of Convolutional Neural Network in Image Classification
Uma Maheswari S., Praburam K. Varadharajan, Shyam Sunder R., Raksheka Rajakumar · International Journal of Computer Applications · 2023
Deep learning algorithms is designed to mimic the function of a brain.In deep learning algorithms, one of the most prominent deep neural networks used for image recognition and segmentation tasks is the Convolutional Neural Network (CNN).In this paper, various types of CNN architectures like VGGNet, AlexNet, ResNet, and LeNet-5 are built and the performances are compared using a publicly available dataset (CIFAR-10).Furthermore, multiple performance optimizers: Root Mean Square Propagation (RMSProp), Adaptive moment estimation (Adam), and Adaptive gradient estimation (Adagrad), are applied for this study.The performance of these five CNN architectures with three optimizers is evaluated in terms of accuracy, specificity, and sensitivity.The experimental results showed that the Inception-V3 model with RMSProp as an optimizer achieved the highest validation accuracy of 92.97% with a misclassification rate of 7.03%.