[PDF][PDF] Chinese painting generation using generative adversarial networks
G Wang, Y Chen, Y Chen - 2017 - guanyangwang.github.io
This project implements different types of Generative Adversarial Networks (GANs) such as
cGANs, DCGANs, WGANs, and our modified WGANs on our own Chinese painting dataset
to recover original paintings from edge maps, and generate realistic-looking paintings. We
also compared the results of different GANs. Empirically, we found that WGANs and our
modified WGANs are more stable and are able to generate images with higher quality. In
particular, the modified WGANs performs well in getting rid of the problem of mode collapse.
cGANs, DCGANs, WGANs, and our modified WGANs on our own Chinese painting dataset
to recover original paintings from edge maps, and generate realistic-looking paintings. We
also compared the results of different GANs. Empirically, we found that WGANs and our
modified WGANs are more stable and are able to generate images with higher quality. In
particular, the modified WGANs performs well in getting rid of the problem of mode collapse.
以上显示的是最相近的搜索结果。 查看全部搜索结果