Open AccessResearch on Fast Image Style Transformation Based on Residual Network
Author(s) -
Nan Xue,
Depeng Jin,
Bing Zhang,
Xiao Yu Wang,
Lintao Yu
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1314/1/012189
Subject(s) - residual , convolution (computer science) , convolutional neural network , computer science , transformation (genetics) , image (mathematics) , image restoration , artificial intelligence , algorithm , artificial neural network , style (visual arts) , point (geometry) , mathematics , geometry , image processing , history , biochemistry , chemistry , archaeology , gene
Image style migration technology refers to the conversion of an image into a similar image to a famous painting style through learning (using a convolutional neural network). The A Neural Algorithm of Artistic Style (NAAS) proposed by Gatys uses the VGG network to design a loss network, and the style migration image is obtained through repeated iterations, and the calculation speed is slow. Li Feifei introduced the residual network on the basis of Gatys, and accelerated the calculation by using the fast connection feature of the residual element. This method has a great improvement in the generation speed but still has room for improvement. This paper proposes improvements for the following two aspects: firstly, the classical residual element structure is adjusted, and the standard convolution is converted into point convolution and deep convolution, which reduces the calculation amount while ensuring the convolution effect; secondly, the loss network is simplified. The fourth and fifth layers in the model are highly consistent in structure, and the style restoration of the two layers is basically the same as the content reconstruction effect. Therefore, the fifth layer is deleted, the redundant parameters are removed, and the parameter amount is reduced. At the same time, the effect of style restoration and content reconstruction is guaranteed.