Journal of Jilin University Science Edition ›› 2021, Vol. 59 ›› Issue (3): 577-586.
Previous Articles Next Articles
LI Erqiang, CHEN Kaijian, ZHOU Yang
Received:
Online:
Published:
Abstract: Aiming at the problems of mode collapse in the application of single texture expansion model to multiple texture expansion, and the user could not control the output texture mode and style, we proposed a novel network for controllable multi-texture extended synthesis and transfer. Firstly, by adding classification training in the discrimination training of generative adversarial network (GAN), the discriminator could distinguish the generated data from the real data, and could further correctly identify the input texture came from which training image, so as to improve the problem of mode collapse. Secondly, in order to achieve the user’s control of texture mode in texture transfer, we modified the generator to a two-stream data input, in which one stream provided structure-guiding features, and the other one provided texture mode features. The final texture image was generated by decoding after fusing the two features. The experimental results show that the proposed model can not only correctly learn the texture modes of multi-texture images with one network, but also has better and controllable texture transfer function.
Key words: texture synthesis, control synthesis, texture transfer, generative adversarial network (GAN)
CLC Number:
LI Erqiang, CHEN Kaijian, ZHOU Yang. Controllable Multi-texture Extended Synthesis and Transfer[J].Journal of Jilin University Science Edition, 2021, 59(3): 577-586.
0 / / Recommend
Add to citation manager EndNote|Reference Manager|ProCite|BibTeX|RefWorks
URL: http://xuebao.jlu.edu.cn/lxb/EN/
http://xuebao.jlu.edu.cn/lxb/EN/Y2021/V59/I3/577
Cited