Real-world paintings are made, by artists, using brush strokes as the rendering primitive to depict semantic content. The bulk of the Neural Style Transfer (NST) is known transferring style using texture patches, not strokes. The output looks like the content image, but some are traced over using the style texture: it does not look painterly. We adopt a very different approach that uses strokes. Our contribution is to analyse paintings to learn stroke families-that is, distributions of strokes based on their shape (a dot, straight lines, curved arcs, etc.). When synthesising a new output, these distributions are sampled to ensure the output is painted with the correct style of stroke. Consequently, our output looks more "painterly" than NST output based on texture. Furthermore, where strokes are placed is an important contributing factor in determining output quality, and we have also addressed this aspect. Humans place strokes to emphasize salient semantically meaningful image content. Conventional NST uses a content loss premised on filter responses that is agnostic to salience. We show that replacing that loss with one based on the language-image model benefits the output through greater emphasis of salient content.

Research has shown that convolutional neural networks for object recognition are vulnerable to changes in depiction because learning is biased towards the low-level statistics of texture patches. Recent works concentrate on improving robustness by applying style transfer to training examples to mitigate against over-fitting to one depiction style. These new approaches improve performance, but they ignore the geometric variations in object shape that real art exhibits: artists deform and warp objects for artistic effect. Motivated by this observation, we propose a method to reduce bias by jointly increasing the texture and geometry diversities of the training data. In effect, we extend the visual object class to include examples with shape changes that artists use. Specifically, we learn the distribution of warps that cover each given object class. Together with augmenting textures based on a broad distribution of styles, we show by experiments that our method improves performance on several cross-domain benchmarks.

Since its inception in 2015, Style Transfer has focused on texturing a content image using an art exemplar. Recently, the geometric changes that artists make have been acknowledged as an important component of style[42], [55], [62], [63]. Our contribution is to propose a neural network that, uniquely, learns a mapping from a 4D array of inter-feature distances to a non-parametric 2D warp field. The system is generic in not being limited by semantic class, a single learned model will suffice; all examples in this paper are output from one model.Our approach combines the benefits of the high speed of Liu et al. [42] with the non-parametric warping of Kim et al. [55]. Furthermore, our system extends the normal NST paradigm: although it can be used with a single exemplar, we also allow two style exemplars: one for texture and another geometry. This supports far greater flexibility in use cases than single exemplars can provide.

State-of-the-art neural style transfer methods have demonstrated amazing results by training feed-forward convolutional neural networks or using an iterative optimization strategy. The image representation used in these methods, which contains two components: style representation and content representation, is typically based on high-level features extracted from pretrained classification networks. Because the classification networks are originally designed for object recognition, the extracted features often focus on the central object and neglect other details. As a result, the style textures tend to scatter over the stylized outputs and disrupt the content structures. To address this issue, we present a novel image stylization method that involves an additional structure representation. Our structure representation, which considers two factors: i) the global structure represented by the depth map and ii) the local structure details represented by the image edges, effectively reflects the spatial distribution of all the components in an image as well as the structure of dominant objects respectively. Experimental results demonstrate that our method achieves an impressive visual effectiveness, which is particularly significant when processing images sensitive to structure distortion, e.g. images containing multiple objects potentially at different depths, or dominant objects with clear structures.