Adversarial Training for Fast Arbitrary Style Transfer
Computers & Graphics(2020)
摘要
•Adversarial training can be applied to learn loss for arbitrary style transfer.•A fast feed forward network can be adversarially trained for arbitrary style transfer.•The stylization level can be controlled by automatic mask module and manual parameters.•Style transfer can be widely applied to multi-domain data.
更多查看译文
关键词
Style transfer,Adversarial networks,Deep learning,Image generation
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络