Stroke-based Neural Painting and Stylization with Dynamically Predicted Painting Region

TL;DR

This paper introduces a dynamic, region-predicting stroke-based neural painting framework that improves boundary consistency and extends to stylization, outperforming existing methods in quality.

Contribution

We propose Compositional Neural Painter, a novel framework that dynamically predicts painting regions and incorporates a differentiable loss for stylization, advancing stroke-based neural rendering.

Findings

Outperforms existing models in neural painting quality

Effectively preserves image structure during stylization

Reduces boundary artifacts in stroke-based rendering

Abstract

Stroke-based rendering aims to recreate an image with a set of strokes. Most existing methods render complex images using an uniform-block-dividing strategy, which leads to boundary inconsistency artifacts. To solve the problem, we propose Compositional Neural Painter, a novel stroke-based rendering framework which dynamically predicts the next painting region based on the current canvas, instead of dividing the image plane uniformly into painting regions. We start from an empty canvas and divide the painting process into several steps. At each step, a compositor network trained with a phasic RL strategy first predicts the next painting region, then a painter network trained with a WGAN discriminator predicts stroke parameters, and a stroke renderer paints the strokes onto the painting region of the current canvas. Moreover, we extend our method to stroke-based style transfer with a novel differentiable distance transform loss, which helps preserve the structure of the input image during stroke-based stylization. Extensive experiments show our model outperforms the existing models in both stroke-based neural painting and stroke-based stylization. Code is available at https://github.com/sjtuplayer/Compositional_Neural_Painter