A Multi-view Images Classification Based on Deep Graph Convolution
2019 6th International Conference on Information Science and Control Engineering (ICISCE)(2019)
摘要
Multi-view images represent the target images from different viewpoints. Reasonable use of complementary information between multiple perspective images helps to improve the accuracy of classification. We propose a novel deep spectral graph convolutional neural network (DSGC) model, extracting the deep characteristics of multi-view images via five spectral graph convolutional layers. Moreover, we propose a new method, which combines the DSGC with the ideal of [1]. The ideal of [1] regards the images of predefined views as latent variables, outputting the category probability matrix. We propose a DualFeature layer, which is used to enhance the ability to extract features. A pair of techniques batch normalization (BN) and dropout are used to improve classification performance. Experimental results show that.
更多查看译文
关键词
Deep Spectral Graph Convolution,DualFeature Layer,Latent Variables,Batch Normalization,dropout,Multiview Images Classification
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要