BriefGPT.xyz
May, 2022
GR-GAN: 逐步细化的文本到图像生成
GR-GAN: Gradual Refinement Text-to-image Generation
HTML
PDF
Bo Yang, Fangxiang Feng, Xiaojie Wang
TL;DR
本文提出了一种渐进式精炼生成对抗网络 (GR-GAN) 用于高效解决生成图像与文本约束的一致性问题;包括 GRG 模块,ITM 模块以及新指标 CMD 的引入,并在实验中表现出了显著的优势。
Abstract
A good
text-to-image model
should not only generate high quality images, but also ensure the consistency between the text and the generated image. Previous models failed to simultaneously fix both sides well. This paper proposes a
→