SketchGNN: Semantic Sketch Segmentation with Graph Neural Networks
This addresses the problem of accurately labeling and segmenting sketches for applications in design and graphics, representing a strong specific gain over existing methods.
The paper tackles semantic segmentation of freehand vector sketches by introducing SketchGNN, a graph neural network that treats sketches as graphs with nodes as sampled points and edges as stroke structure, achieving state-of-the-art improvements of 11.2% in pixel-based and 18.2% in component-based metrics on the SPG dataset.
We introduce SketchGNN, a convolutional graph neural network for semantic segmentation and labeling of freehand vector sketches. We treat an input stroke-based sketch as a graph, with nodes representing the sampled points along input strokes and edges encoding the stroke structure information. To predict the per-node labels, our SketchGNN uses graph convolution and a static-dynamic branching network architecture to extract the features at three levels, i.e., point-level, stroke-level, and sketch-level. SketchGNN significantly improves the accuracy of the state-of-the-art methods for semantic sketch segmentation (by 11.2% in the pixel-based metric and 18.2% in the component-based metric over a large-scale challenging SPG dataset) and has magnitudes fewer parameters than both image-based and sequence-based methods.