File size: 2,901 Bytes
ab4bf28 674a765 ab4bf28 674a765 ab4bf28 674a765 ab4bf28 674a765 ab4bf28 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 | ---
tags:
- svg
- vector-graphics
- font-generation
- transformer
- variational-autoencoder
- generative-model
---
<h1 align="center">DesigNet: Learning to Draw Vector Graphics as Designers Do</h1>
<p align="center">
<a href="https://colab.research.google.com/github/TomasGuija/DesigNet/blob/main/demo_designet_inference.ipynb">
<img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open in Colab">
</a>
<a href="https://arxiv.org/abs/2604.06494">
<img src="https://img.shields.io/badge/arXiv-2604.06494-b31b1b.svg" alt="arXiv">
</a>
<a href="https://doi.org/10.1016/j.cag.2026.104627">
<img src="https://img.shields.io/badge/DOI-10.1016%2Fj.cag.2026.104627-blue.svg" alt="Journal DOI">
</a>
</p>
<p align="center">
<img
src="https://raw.githubusercontent.com/TomasGuija/DesigNet/main/docs/imgs/overview.jpg"
width="650"
alt="Overview of the DesigNet architecture"
>
</p>
- **Code:** [github.com/TomasGuija/DesigNet](https://github.com/TomasGuija/DesigNet)
- **Published article:** [Computers & Graphics](https://doi.org/10.1016/j.cag.2026.104627)
- **Preprint:** [arXiv:2604.06494](https://arxiv.org/abs/2604.06494)
## Overview
AI-driven content generation has made remarkable progress in recent years. However, neural networks and human designers operate in fundamentally different ways, making collaboration between them challenging. We address this gap for Scalable Vector Graphics (SVG) by equipping neural networks with tools commonly used by designers, such as axis alignment and explicit continuity control at command junctions. We introduce DesigNet, a hierarchical Transformer-VAE that operates directly on SVG sequences with a continuous command parameterization. Our main contributions are two differentiable modules: a continuity self-refinement module that predicts \\(C^{0}\\), \\(G^{1}\\), and \\(C^{1}\\) continuity for each curve point and enforces it by modifying Bézier control points, and an alignment self-refinement module with snapping capabilities for horizontal or vertical lines. DesigNet produces editable outlines and achieves competitive results against state-of-the-art methods, with notably higher accuracy in continuity and alignment. These properties ensure that the outputs are easier to refine and integrate into professional design workflows.
## Citation
```bibtex
@article{GUIJAVALIENTE2026104627,
title = {DesigNet: Learning to draw vector graphics as designers do},
journal = {Computers & Graphics},
volume = {137},
pages = {104627},
year = {2026},
issn = {0097-8493},
doi = {https://doi.org/10.1016/j.cag.2026.104627},
url = {https://www.sciencedirect.com/science/article/pii/S0097849326000981},
author = {Tomas Guija-Valiente and Iago Suárez},
keywords = {Scalable vector graphics, Font generation, Transformer, Variational Autoencoder, Generative AI},
}
``` |