CLIPVG: Text-Guided Image Manipulation Using Differentiable Vector Graphics

Authors

  • Yiren Song Shanghai Jiao Tong University Netease Games AI Lab
  • Xuning Shao Netease Games AI Lab
  • Kang Chen NetEase Games AI Lab
  • Weidong Zhang Netease Games AI Lab
  • Zhongliang Jing Shanghai Jiao Tong University
  • Minzhe Li Shanghai Jiao Tong University

DOI:

https://doi.org/10.1609/aaai.v37i2.25326

Keywords:

CV: Applications, CV: Language and Vision, ML: Unsupervised & Self-Supervised Learning

Abstract

Considerable progress has recently been made in leveraging CLIP (Contrastive Language-Image Pre-Training) models for text-guided image manipulation. However, all existing works rely on additional generative models to ensure the quality of results, because CLIP alone cannot provide enough guidance information for fine-scale pixel-level changes. In this paper, we introduce CLIPVG, a text-guided image manipulation framework using differentiable vector graphics, which is also the first CLIP-based general image manipulation framework that does not require any additional generative models. We demonstrate that CLIPVG can not only achieve state-of-art performance in both semantic correctness and synthesis quality, but also is flexible enough to support various applications far beyond the capability of all existing methods.

Downloads

Published

2023-06-26

How to Cite

Song, Y., Shao, X., Chen, K., Zhang, W., Jing, Z., & Li, M. (2023). CLIPVG: Text-Guided Image Manipulation Using Differentiable Vector Graphics. Proceedings of the AAAI Conference on Artificial Intelligence, 37(2), 2312-2320. https://doi.org/10.1609/aaai.v37i2.25326

Issue

Section

AAAI Technical Track on Computer Vision II