Mostly we're using the _nnls function for blendshape fitting, but just for completeness, this function is now changed to use only Eigen as well.

Convenience function so they can easily be multiplied with a blendshapes matrix.

All matrix operations and solving the linear system is now done with Eigen matrices.
We use colPivHouseholderQr for now to solve the linear system, which doesn't have any requirements on the matrix.

When switching the shape and blendshape fitting to use the new Eigen models, these matrices weren't correctly mapped to cv::Mat. Eigen uses col-major storage order by default, and OpenCV row-major - so we now convert to a row-major Eigen matrix before mapping to a cv::Mat.
This caused the fitting to produce faulty/meaningless results since the switch to Eigen.
The output of the linear shape and blendshape fitting functions are now again identical before and after the switch to Eigen. (after 50 iterations, the result is marginally different, most likely because of small rounding differences.)

* ImageCost throws on instantiation if the given model doesn't contain a colour model
* Changed the default parameter of fit-model-ceres to sfm_shape_3448, which will need to be obtained
* Additionally the app checks explicitly for has_color_model() and will exit if it hasn't got one.

This fixes a crash - it should obviously return a matrix, not a vector.

This seems to be unused (uninstantiated) code, and only clang complains, but not MSVC & gcc.

This removes the last usage of OpenCV in PcaModel.

It's no longer necessary, since the RNG is no longer part of PcaModel.

This makes the intent clearer - it does not return the i-th basis vector (which one could easily assume, if not reading the function parameter name or documentation).
Also added a note expressing this.
It also leaves room for a future function that actually does return the i-th basis.

And compute the rescaled basis when loading a model.

Adjusted all files that use it (hopefully).

This is a breaking change: If you're currently using the PcaModel constructor and giving it a PCA basis matrix, your code will (probably silently) break: The function now expects an orthonormal basis matrix, whereas before, it expected a rescaled basis matrix. However it makes more sense for the function to take an orthonormal basis matrix.

This should clear up confusion and make the descriptions good and unambiguous.

- orthonormal = the old "unnormalised"
- rescaled = the old "normalised".

Also changed PcaModel's getter for the basis accordingly, and updated the documentation.

'did' was only really an external thing, so naming it "SFM" is more clear!

Renamed all references to "ibug2did" to "ibug_to_sfm".

Added explicit casts to glm::vec3 from glm::vec4 as it was causing
build errors with clang 3.9.1 on OS X.

Changed assignment operation to initialization for ray_origin.

Eigen only introduced rsqrt() in 3.3. Eventually, we'll switch back to rsqrt(), or even include Eigen directly under 3rdparty/.

For some reason, not all changes got staged in the last commit. This is the full set of changes.

Actually, we still read into an OpenCV matrix - should change that in the future to directly read into an Eigen::Matrix.

This is a big step towards being not requiring OpenCV anymore as a dependency.
All models will have to be converted and re-downloaded.

* PcaModel: Removed the RNG from the class, and added a RNG parameter to the draw_sample method
* PcaModel: Added a get_eigenvalues() accessor that returns all eigenvalues
* MorphableModel: Changed CEREAL_CLASS_VERSION from 0 to 1 (and added a version check to the serialize method)
* MorphableModel: Changed texture_coordinates from cv::Vec2f to array<double, 2>
* Added binary Eigen::Matrix serialisation for cereal
* Changed Blendshape to Eigen::Vector as well
* Adjusted blendshape_fitting.hpp accordingly - and it now takes most parameters by const&
* Adjusted fitting.hpp accordingly - now also takes most parameters by const&
* Adjusted linear_shape_fitting.hpp: It's not totally converted, still using cv::Mat for the core algorithm

It hasn't had a method to return a Mesh in quite a while I think.

Hopefully this is not an issue anymore.

These are required for extract_texture in Matlab.
Maybe the better strategy would be to add a constructor to RenderingParameters and not have all these setters.

Replaced all instances of old Vertex class with Vertex<float>, so now there's only one vertex class.

Forgot to add in earlier commit - this should fix compile error.
Also added cv:: in Rasterizer to fix second compile error