Implementation of method described in the paper "Mesh-Based Inverse Kinematics" by Sumner et al. 2005. This method learns the features of a mesh through example poses and uses these features to generate new meaningful poses that satisfy user inputed position constraints.
Techniques used: mesh processing, linear algebra, optimization
Run 'meshik.py' to see an example.
Todo
- Refactor/organize/comment code
- Implement rodrigues exponential map
- Include different ways of solving (ex. interpolation without constraints)
- Vectorize nonlinear combination of feature vectors
- Produce some cool examples
- Add a nice explanation of the method