Classification of Famous Paintings Based on Convolutional Neural Network
Main Article Content
Abstract
Neural network is a method to realize machine learning. In recent years, it has achieved good performance in completing many tasks. Convolutional Neural Network (CNN) is one of the most important algorithms in neural networks, and it performs well in the field of image classification. This paper studies the classification of famous paintings based on CNN. Firstly, I analyze the influence of network depth and iteration number on classification accuracy. Then, I propose an optimal network for paintings classification. To solve the problem of over-fitting, I use a data augmentation method. The processed data set is used as the input to the neural network, and then the input through convolution layers, subsampling layers, and two fully connected layers connected by the a dropout layer. Besides, I use the combination of Sigmoid and LeakyReLU functions as the activation function. Finally, compared with the traditional deep learning methods, the model proposed in this paper achieves better results with the accuracy of about 0.82.