/Hyperspectral-Inverse-Skinning

Code and data for "Hyperspectral Inverse Skinning" by Songrun Liu, Jianchao Tan, Zhigang Deng, Yotam Gingold in Computer Graphics Forum 2020

Primary LanguageJavaScriptMIT LicenseMIT

InverseSkinning

This is the repository for the code and data used in the paper Hyperspectral Inverse Skinning by Songrun Liu, Jianchao Tan, Zhigang Deng, Yotam Gingold in Computer Graphics Forum 2020.

To generate the weights and transformations used in the paper, run the scripts:

results_songrun_final_run_sh/*/run.sh

For Table 3:

  • cat-poses
  • chickenCrossing
  • elephant-gallop
  • elephant-poses
  • face-poses
  • horse-collapse
  • horse-gallop
  • horse-poses
  • lion-poses
  • pcow
  • pdance
  • pjump

and compare to the data in SSD_unconstrained.

For Table 2:

  • crane
  • elasticCow
  • elephant
  • horse
  • samba

and compare to existing output data (see below).

For Table 5:

  • cylinder
  • cube
  • cheburashka
  • wolf
  • cow

Various comparison scripts can be found as compare*.py. Various other scripts can be found as test*.sh. Some examples can be found in COMMANDS.md.

The error and running time for Le and Deng's [2012] output comes from Table 3 of their paper. All recovered OBJs for our method, and a subset of the computed OBJs for Le and Deng [2012] and Luo et al. [2019] can be directly downloaded from the project page.

Dependencies:

  • numpy (e.g. pip install numpy)
  • scipy (e.g. pip install scipy)
  • MOSEK (download the software and license, cd mosek/8/tools/platform/osx64x86/python/3/; python setup.py install)
  • cvxopt You need an unreleased version with import MOSEK binding fixes: CVXOPT_BUILD_GLPK=1 pip install git+https://github.com/cvxopt/cvxopt/
  • (viewer only) libigl(e.g. git clone https://github.com/libigl/libigl.git --recursive)
  • (viewer only) CGAL (e.g. brew install cgal)
  • (viewer only) eigen (e.g. brew install eigen)
  • (viewer only) GLFW3 (e.g. brew install glfw3)

Data

See models/README.md.

Compile the viewer

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make

Running the GUI

View a set of poses with random coloring on each handle's influencing region.

./viewer ../models/animal/poses-1/animal-1.obj ../models/animal/poses-1/animal-2.obj ../models/animal/poses-1/animal-3.obj ../models/animal-estimated.DMAT

Defunct weight estimation (no longer used)

Assuming a set of poses and corresponding handle(bone)'s transformation matrices(Tmat, #handle*4-by-3) are given, Estimate each vertex's weights.

./estimate_weights ../models/animal/poses-*/animal-*.obj ../models/animal/poses-*/animal-*.Tmat --weight ../models/animal/animal.DMAT

Test the minimum volume enclosure simplex

On the minimum volume simplex enclosure problem for estimating a linear mixing model (https://link.springer.com/article/10.1007/s10898-012-9876-5).

python simplex_hull.py models/cube4/poses-1