论文标题
多属性引导绘画一代
Multi-Attribute Guided Painting Generation
论文作者
论文摘要
可控的绘画生成在图像样式中起关键作用。当前，样式转移的控制方式受到基于示例的参考或随机的单热矢量指南的约束。很少有工作重点是将绘画作为控制条件的固有特性，例如艺术家，流派和时期。在这种情况下，我们提出了一个新颖的框架，该框架采用了绘画中的多个属性来控制风格化的结果。不对称的循环结构可以保留保真度，并与样式保存和属性回归损失相关联，以保持域之间颜色和纹理的独特区别。几个定性和定量结果证明了多个属性组合的影响并实现令人满意的性能。
Controllable painting generation plays a pivotal role in image stylization. Currently, the control way of style transfer is subject to exemplar-based reference or a random one-hot vector guidance. Few works focus on decoupling the intrinsic properties of painting as control conditions, e.g., artist, genre and period. Under this circumstance, we propose a novel framework adopting multiple attributes from the painting to control the stylized results. An asymmetrical cycle structure is equipped to preserve the fidelity, associating with style preserving and attribute regression loss to keep the unique distinction of colors and textures between domains. Several qualitative and quantitative results demonstrate the effect of the combinations of multiple attributes and achieve satisfactory performance.
