It fits pose and shape, uses front-facing contour as well as occluding contour fitting. It can iterate (alternate) the pose and shape fitting.
A simpler overload exists with less parameters and reasonable defaults.

Also removed unnecessary model_points vector

The function was often already called with degrees, not radians, so this resulted in converting to degree twice. It didn't matter because only a comparison of <0 and >0 was done. But it's fixed and clearly documented now that the function expects the yaw angle in degrees.

This contains a NNLS solver, which we use to solve for the blendshape coefficients.

We have two of these functions now, one for glm::vec3 and one for cv::Vec3f. At some point, this needs clean-up!

Added a description of the file to share/readme

Using nanoflann's kd-tree for fast searching

Fixed previously uninitialised screen_width/height in constructor from ScaledOrthoProjectionParameters

These changes are for the closest edge fitting.

* Function to compute the mesh contour vertices
* nanoflann adaptor for vector<Eigen::Vector2d> (it should work with glm::vec2 too actually)

clang and gcc complained about this, while VS2015 accepted it. It turns out this is the much better way to do it!

Added SOP algorithm to features, added Python bindings section, updated GLM path and added note to clone with --recursive.