基于并行卷积神经网络的图像美学分类

Image aesthetic classification based on parallel convolutional neural network

针对传统的图像美学分类大多基于手动提取图片的美学特征,利用神经网络等方法对图像进行分类,存在分类效果不佳的问题,提出一种基于深度学习的并行卷积神经网络算法,改进图像美学分类的方法。从图像的不同角度出发,自动提取有用的图像美学特征,提高对图像美学的分类效果。实验结果表明,与其它算法实验结果相对比,所提算法增加了图像美学分类的准确率,有一定的实用性。

Aesthetic classification of traditional image aesthetics is mostly based on manual extraction of aesthetics features of images,and using neural networks and other methods to classify the images leads to poor classification results.To solve the problems,aparallel learning algorithm based on convolution neural network was presented to improve the method of image aesthetics classification.Useful image aesthetic features were automatically extracted from different perspectives of the image,thereby improving the classification effects of image aesthetics.The results show that compared with the experimental results of other algorithms,the accuracy of image aesthetics classification is increased,the proposed method shows some practical value.