Code & Data

Human face modeling

• Multilinear autoencoders (WACV 2018): multilinear model learned from a large database with missing and noisy data efficiently using a novel autoencoder architecture that fixes the decoder to a tensor-based model and code to use this model.
• Robust multilinear model learning framework for 3D faces (CVPR 2016): code to learn a multilinear model of high quality from a database with missing data, partial data, and erroneous correspondence information. On this page, the optimized face models can also be downloaded.
• Multilinear correspondence optimization for 3D faces (ICCV 2015): code to jointly optimize a multilinear model and the registration of the 3D scans used for training. On this page, the optimized face models can also be downloaded.
• Statistical 3D face models (ECCV 2014, CVIU 2014, CVIU 2015): four high-quality statistical 3D shape models of human faces along with code to fit the models to noisy input scans.

Human body modeling

• Statistical 3D body models (PR 2017): family of expressive 3D human body shape models and tools for human shape space building, manipulation and evaluation.
• Dressed human bodies in motion dataset (ECCV 2016): benchmark containing 6 different subjects performing 3 motions in 3 different clothing styles each.

3D shape modeling

• Robust partial isometric correspondence computation of 3D shapes (GMOD 2014): code to compute dense intrinsic point-to-point correspondences between two 3D models that is particularly robust to topological noise.
• Partial intrinsic symmetry computation on 3D shapes (NORDIA 2014): code to compute partial intrinsic symmetries of a given scale on a single 3D shape. Both continuous and discrete symmetries are computed.