BriefGPT.xyz
May, 2017
姿势引导的人像图像生成
Pose Guided Person Image Generation
HTML
PDF
Liqian Ma, Xu Jia, Qianru Sun, Bernt Schiele, Tinne Tuytelaars...
TL;DR
本论文提出了新颖的姿势引导人物生成网络(PG2),该网络可以基于人的图像和新颖的姿态来合成人体图像。PG2的生成框架明确利用了姿势信息,并包含两个关键阶段:姿势整合和图像细化。实验表明,该模型可以生成高质量、具有令人信服细节的人体图像。
Abstract
This paper proposes the novel
pose guided person generation network
(PG$^2$) that allows to synthesize
person images
in arbitrary poses, based on an image of that person and a novel pose. Our generation framework
→