Please wait a minute...
Vis Inf
Vis Inf  0, Vol. Issue (): 59-69    DOI: 10.1016/j.visinf.2019.06.001
论文     
Hair-GAN：采用生成对抗网络从单幅图像恢复3D头发结构
Meng Zhang，Youyi Zheng
State Key Lab of CAD&CG, Zhejiang University, Hang Zhou 310058, China
Hair-GAN: Recovering 3D hair structure from a single image using generative adversarial networks
Meng Zhang，Youyi Zheng
State Key Lab of CAD&CG, Zhejiang University, Hang Zhou 310058, China
 全文: PDF 
摘要： 本文提出了一种生成对抗网络的架构Hair-GAN，可从单个图像中恢复3D头发结构。该网络意在建立从2D头发图像到3D头发结构的参数化变换。本文采用3D体素场来表示3D头发结构，它不仅记录了头发缕所占用的空间而且记录了其方向。给定一幅头发图像，我们首先将其与一人头的泡沫模型配准，然后提取一组记录了头发方向信息的2D图像，连同泡沫模型深度图一起输入到Hair-GAN中。 通过发生器网络可计算出头发的3D体素场，为最终头发合成提供结构指导。建模结果不仅保持了与输入图像中头发风格的相似性，而且从其他视线方向看具有许多生动的细节。通过模拟多种发型并与现有技术进行比较，证明了本文的方法的有效性。 
关键词： 单视图头发建模;  三维体积结构;  深度学习;  生成对抗网络    
Abstract: We introduce Hair-GAN, an architecture of generative adversarial networks, to recover the 3D hair structure from a single image. The goal of our networks is to build a parametric transformation from 2D hair maps to 3D hair structure. The 3D hair structure is represented as a 3D volumetric field which encodes both the occupancy and the orientation information of the hair strands. Given a single hair image, we first align it with a bust model and extract a set of 2D maps encoding the hair orientation information in 2D, along with the bust depth map to feed into our Hair-GAN. With our generator network, we compute the 3D volumetric field as the structure guidance for the final hair synthesis. The modeling results not only resemble the hair in the input image but also possesses many vivid details in other views. The efficacy of our method is demonstrated by using a variety of hairstyles and comparing with the prior art.
Key words: Single-view hair modeling    3D volumetric structure    Deep learning    Generative adversarial networks
出版日期: 2019-06-27
通讯作者: Youyi Zheng     E-mail: youyizheng@zju.edu.cn
服务　　
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章　　
Meng Zhang
Youyi Zheng

引用本文:

Meng Zhang, Youyi Zheng. Hair-GAN: Recovering 3D hair structure from a single image using generative adversarial networks. Vis Inf, 0, (): 59-69.

链接本文:

http://www.zjujournals.com/vi/CN/10.1016/j.visinf.2019.06.001    或    http://www.zjujournals.com/vi/CN/Y0/V/I/59

[1] Georgios Lampropoulos, Euclid Keramopoulos, Konstantinos Diamantaras. 使用深度学习、语义网络和知识图谱提高增强现实的功能 ——综述 [J]. Vis Inf, 2020, 4(1): 32-42.
[2] Meng Zhang, Youyi Zheng. Hair-GAN：采用生成式对抗网络从单幅图像恢复3D头发结构[J]. Vis Inf, 2019, 3(2): 102-112.
[3] Rulei Yu, Lei Shi. 深度学习可视化综述：面向用户群体分类 [J]. Vis Inf, 2018, 2(3): 147-154.