Stylized Neural Painting


Zhengxia Zou (1);  Tianyang Shi (2);  Shuang Qiu (1);  Yi Yuan (2);  Zhenwei Shi (3)

(1) University of Michigan, Ann Arbor;  (2) NetEase Fuxi AI Lab;  (3) Beihang University

  [Preprint]       [Code]       [Colab]



This paper proposes an image-to-painting translation method that generates vivid and realistic painting artworks with controllable styles. Different from previous image-to-image translation methods that formulate the translation as pixel-wise prediction, we deal with such an artistic creation process in a vectorized environment and produce a sequence of physically meaningful stroke parameters that can be further used for rendering. Since a typical vector render is not differentiable, we design a novel neural renderer which imitates the behavior of the vector renderer and then frame the stroke prediction as a parameter searching process that maximizes the similarity between the input and the rendering output. We explored the zero-gradient problem on parameter searching and propose to solve this problem from an optimal transportation perspective. We also show that previous neural renderers have a parameter coupling problem and we re-design the rendering network with a rasterization network and a shading network that better handles the disentanglement of shape and color. Experiments show that the paintings generated by our method have a high degree of fidelity in both global appearance and local textures. Our method can be also jointly optimized with neural style transfer that further transfers visual style from other images.




In the following we show some stylized paintings generated by our method. Our method can generate vivid paintings with a high degree of realism and artistic sense in terms of both global visual appearance and local texture fidelity. In (d), we also show some highly abstract tape arts of cartoon characters generated by our method. Can you guess who they are?.



Since we frame our stroke prediction under a parameter searching paradigm, our method naturally fits the neural style transfer framework. In the following we show some of our painting results as well as their style transfer results.

 





Our method can also be used for creating 8-bit graphic artworks. In the following we show the artworks created by our method and those by a famous artist Adam Lister. Adam Lister is an American-born artist and painter. In his pixelated paintings, he explores iconic images burned into the psyche from exposure to art history and pop culture. The manual artworks are from his personal gallery website.



Since our painting results are generated with a vector format. We can render them at any resolutions. Here we show two example results rendered at a 1024x1024 pixel resolution.




@inproceedings{zou2020stylized,
    title={Stylized Neural Painting},
      author={Zhengxia Zou and Tianyang Shi and Shuang Qiu and Yi Yuan and Zhenwei Shi},
      year={2020},
      eprint={2011.08114},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}