Learning from Multi-domain Artistic Images for Arbitrary Style Transfer
This work addresses the challenge of generating high-quality stylized images efficiently for applications in digital art and media, though it is incremental with novel components like a mask module.
The authors tackled the problem of arbitrary style transfer by proposing a fast feed-forward network that generates stylized images for unseen content and style pairs, using adversarial networks trained on multi-domain artistic images and achieving qualitative and quantitative improvements over recent methods.
We propose a fast feed-forward network for arbitrary style transfer, which can generate stylized image for previously unseen content and style image pairs. Besides the traditional content and style representation based on deep features and statistics for textures, we use adversarial networks to regularize the generation of stylized images. Our adversarial network learns the intrinsic property of image styles from large-scale multi-domain artistic images. The adversarial training is challenging because both the input and output of our generator are diverse multi-domain images. We use a conditional generator that stylized content by shifting the statistics of deep features, and a conditional discriminator based on the coarse category of styles. Moreover, we propose a mask module to spatially decide the stylization level and stabilize adversarial training by avoiding mode collapse. As a side effect, our trained discriminator can be applied to rank and select representative stylized images. We qualitatively and quantitatively evaluate the proposed method, and compare with recent style transfer methods. We release our code and model at https://github.com/nightldj/behance_release.