StarVector: Generating Scalable Vector Graphics Code from Images and Text

TL;DR

StarVector is a multimodal large language model that generates compact, semantically rich SVG graphics from images and text by understanding image semantics and utilizing diverse SVG primitives.

Contribution

It introduces StarVector, a novel SVG generation model that directly operates in SVG code space and leverages a new dataset and benchmark for improved vector graphic synthesis.

Findings

StarVector outperforms previous methods in generating compact SVGs.

The SVG-Stack dataset enables better generalization across vectorization tasks.

SVG-Bench provides a comprehensive evaluation framework for SVG generation.

Abstract

Scalable Vector Graphics (SVGs) are vital for modern image rendering due to their scalability and versatility. Previous SVG generation methods have focused on curve-based vectorization, lacking semantic understanding, often producing artifacts, and struggling with SVG primitives beyond path curves. To address these issues, we introduce StarVector, a multimodal large language model for SVG generation. It performs image vectorization by understanding image semantics and using SVG primitives for compact, precise outputs. Unlike traditional methods, StarVector works directly in the SVG code space, leveraging visual understanding to apply accurate SVG primitives. To train StarVector, we create SVG-Stack, a diverse dataset of 2M samples that enables generalization across vectorization tasks and precise use of primitives like ellipses, polygons, and text. We address challenges in SVG evaluation, showing that pixel-based metrics like MSE fail to capture the unique qualities of vector graphics. We introduce SVG-Bench, a benchmark across 10 datasets, and 3 tasks: Image-to-SVG, Text-to-SVG generation, and diagram generation. Using this setup, StarVector achieves state-of-the-art performance, producing more compact and semantically rich SVGs.