← ArchivePaper2020
FaceScape: A Large-Scale High Quality 3D Face Dataset and Detailed Riggable 3D Face Prediction
Haotian Yang, Hao Zhu, Yanru Wang, Mingkai Huang, Qiu Shen, Ruigang Yang, Xun Cao
18,760 pore-level textured 3D faces from 938 subjects with 20 expressions, with a riggable 3D face prediction method from single images.
Abstract
In this paper, we present a large-scale detailed 3D face dataset, FaceScape, and propose a novel algorithm that is able to predict elaborate riggable 3D face models from a single image input. FaceScape dataset provides 18,760 textured 3D faces, captured from 938 subjects and each with 20 specific expressions. The 3D models contain the pore-level facial geometry that is also processed to be topologically uniformed. These fine 3D facial models can be represented as a 3D morphable model for rough shapes and displacement maps for detailed geometry. Taking advantage of the large-scale and high-accuracy dataset, a novel algorithm is further proposed to learn the expression-specific dynamic details using a deep neural network. The learned relationship serves as the foundation of our 3D face prediction system from a single image input. Different than the previous methods, our predicted 3D models are riggable with highly detailed geometry under different expressions. The unprecedented dataset and code will be released to public for research purpose.
How to read this
- Category
- Dataset plus method: 3D face dataset and riggable prediction
- Contributions
- FaceScape, a large dataset of pore-level textured, topologically uniform 3D faces (18,760 models, 938 subjects, 20 expressions)
- A representation as a 3D morphable model for coarse shape plus displacement maps for fine detail
- A network that learns expression-specific dynamic details to predict a riggable, detailed 3D face from a single image
- Context
- Extends the tradition of 3D facial expression databases and morphable-model fitting, positioned relative to Cao et al.'s FaceWarehouse.Builds on: FaceWarehouse: A 3D Facial Expression Database for Visual Computing
- Correctness
- The riggable single-image prediction rests on the dataset's scale and accuracy and on the coarse-shape-plus-displacement split; readers should remember that subject diversity is bounded by the 938 captured individuals and that single-image inference of fine detail is an inferred reconstruction, not a measurement.
- Clarity
- Accessible; the dataset and goals read clearly on a first pass, the detail-learning network needs a second pass.
- How to read it
- First pass to assess dataset scope and licensing for your use; second pass on the displacement-map detail learning and the single-image prediction pipeline if you intend to fit or train on it.
Related work
- Real-Time High-Fidelity Facial Performance Capture 2015 / SIGGRAPH
- Single-Shot High-Quality Facial Geometry and Skin Appearance Capture 2020 / SIGGRAPH
- Advances for Digital Humans in VFX Production at Goodbye Kansas Studios 2022 / SIGGRAPH
- Driving High-Resolution Facial Scans with Video Performance Capture 2015 / SIGGRAPH
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →