including implementations of shell deformation energy cost, gradient, and hessian derivations;
including implementation of elastic average and geodesic average;
including geodesic interpolation and extrapolation;
including implementation of objective function for model fitting and mesh editing;
c3d_data:
includes *.c3d type motion capture data selected from CMU MoCap repo and MPI MoSh website
faust:
includes 10 sample meshes taken from MPI FAUST dataset
precomputed:
includes some precomputed model files or data files for the usage of running some included demos
c3d_matlab:
includes scripts to load and read c3d files in matlab
others:
includes supportive codes for mesh loading and saving, etc
Yes. We provide several demo scripts which can be used readily to reproduce the results as shown in our paper. With just a few modifications, you can work on your own data.
- filename: demo_GCA.m this code implements the core part of our paper, it produces the average shape and a set of principal variations shapes, given a couple of meshes; the current version only supports working on low resolution meshes; stand-alone deformation transfer codes may be provided in the future;
requirements: a set of decimated meshes with the same topology
- filename: demo_Recon.m this code implements shape reconstruction using the trained PGA model
requirements: 1. pre-computed PGA model 2. test data with the same topology as training data
- filename: demo_fitting_to_markers_.m this code implements fitting the (human body) model to motion capture data (c3d file)
requirements: 1. pre-computed PGA model 2. motion capture data in c3d format
- filename: demo_mesh_edit.m this code implements soft-constraint mesh editing (hard-constraint may be released in the future)
requirements: 1. pre-computed PGA model 2. model correspondence indices to handles, and new handle positions 3. (optional) training data of the model to be used as appropriate initialization