Abstract:

The process of classifying facial data is one of the deep learning topics that has been studied extensively. In this regard, convolutional neural networks are among the most preferred deep learning applications because they are easy to implement and give successful results. On the other hand, the pooling layer in convolutional neural networks causes information loss in the data. In addition, convolutional neural networks perform the training process by ignoring the relative states of the components in the data. As a solution to this situation, capsule networks are one of the recommended deep learning methods. In this study; A data set consisting of 200 data for each of the 5 people and a total of 1000 image data was prepared. With the prepared data set, each of the AlexNet, Vgg16, Vgg19, ResNet50 algorithms, which are frequently used convolutional neural network methods, and the proposed capsule network method were used for training. As a result of the study, the performances of the convolutional neural network methods and the performances of the capsule networks method were compared.

Keywords: