Head generation with diverse identities is an important task in computer vision and computer graphics, widely used in multimedia applications. However, current full head generation methods require a large number of 3D scans or multi-view images to train the model, resulting in expensive data acquisition cost. To address this issue, we propose Head3D, a method to generate full 3D heads with limited multi-view images. Specifically, our approach first extracts facial priors represented by tri-planes learned in EG3D, a 3D-aware generative model, and then proposes feature distillation to deliver the 3D frontal faces into complete heads without compromising head integrity. To mitigate the domain gap between the face and head models, we present dual-discriminators to guide the frontal and back head generation, respectively. Our model achieves cost-efficient and diverse complete head generation with photo-realistic renderings and high-quality geometry representations. Extensive experiments demonstrate the effectiveness of our proposed Head3D, both qualitatively and quantitatively.

本文从生成式模型的视角对三维数字人技术进行梳理, 首先整体介绍生成式三维数字人的建模流 程, 分解出其中的 3 个主要步骤 (第 2 节). 然后分别介绍数字人表示方法 (第 3 节), 数字人渲染方 法 (第 4 节), 以及模型的学习方式 (第 5 节). 之后列举了数字人的一些典型应用 (第 6 节), 最后指 出现有挑战并对未来进行展望 (第 7 节). 已有一些文献对数字人的某类建模或渲染方法进行总结, 如 3DMM 模型 [4] 、人脸重建 [5] 、人体重建 [6, 7] 、三维渲染 [8, 9] 等, 与这些文献不同, 本文旨在从生成式模 型的视角对三维数字人 (人脸及人体) 技术进行全面回顾, 重点介绍基于神经网络的数字人研究方法, 梳理其技术发展趋势及典型应用场景, 让读者能够较为全面地了解数字人的生成技术. 值得注意的是, 除了人脸与人体之外, 头发、手、服饰、骨架等模型同样也属于数字人的研究范畴, 但并非本文的主要 关注对象, 相关内容将在 7.1 和 7.2 小节予以讨论.1) https://www.nlm.nih.gov/research/visible/visible human.html.2) 图 1(a) 左: "Matilda" (https://skfb.ly/6zGMG) by nicolekeane under CC Attribution License; 右: "Girl speedsculpt" (https://skfb.ly/68RCQ) by mrrobot under CC Attribution License.3) 图 1(b) 左: "Red Skull (Rigged)" (https://skfb.ly/oFo7M) by CAPTAAINR under CC Attribution License; 右:"Amazing Spider-Man 2 Movie" (https://skfb.ly/orRTo) by CGI DUDE under CC Attribution License. 4) 图 1(c) 左: Emily [1] ; 右: Obama [1] .