We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: StyleRig: Rigging StyleGAN for 3D Control over Portrait Images

Abstract: StyleGAN generates photorealistic portrait images of faces with eyes, teeth, hair and context (neck, shoulders, background), but lacks a rig-like control over semantic face parameters that are interpretable in 3D, such as face pose, expressions, and scene illumination. Three-dimensional morphable face models (3DMMs) on the other hand offer control over the semantic parameters, but lack photorealism when rendered and only model the face interior, not other parts of a portrait image (hair, mouth interior, background). We present the first method to provide a face rig-like control over a pretrained and fixed StyleGAN via a 3DMM. A new rigging network, RigNet is trained between the 3DMM's semantic parameters and StyleGAN's input. The network is trained in a self-supervised manner, without the need for manual annotations. At test time, our method generates portrait images with the photorealism of StyleGAN and provides explicit control over the 3D semantic parameters of the face.
Comments: CVPR 2020 (Oral). Project page: this https URL
Subjects: Computer Vision and Pattern Recognition (cs.CV); Graphics (cs.GR)
Cite as: arXiv:2004.00121 [cs.CV]
  (or arXiv:2004.00121v2 [cs.CV] for this version)

Submission history

From: Ayush Tewari [view email]
[v1] Tue, 31 Mar 2020 21:20:34 GMT (9064kb,D)
[v2] Sat, 13 Jun 2020 09:40:33 GMT (9058kb,D)

Link back to: arXiv, form interface, contact.