1. Goal
3D human face model generation without any special equipment such as a range scanner as easily as
possible.

2. Abstract
We propose a method for modeling 3D face from a facial photograph using prior knowledge of facial
geometries. Based on the deformable face model and the facial geometry distribution model trained
from approximately 1,000 peoples’ face models, an individual facial 3D geometry can be estimated. As
a result, our system can automatically generate an individual 3D face model with the plausible
facial geometry.

3. Features
The proposed method can generate an individual face model in 1.2[s] with in 2.1[mm] estimation
error. We believe that our method contribute to generate a CG character instantly and automatically
with personal characteristics taking part in a TV game or user attainable entertainment.