Séverin Lemaignan

@Mizerb ok, then it's not due to an empty image... interesting...

yes ;-)

OpenGL `transformFeedback` is probably the way to go.

I'm perfectly fine with using this convention (previous one look like a bug to me), but we probably need first to fix #19, I guess...

What about keeping the one with the smallest perimeter? this should be the one the closest to the actual marker.

One option that could help mitigate: use thresholding instead of Canny: l.90 of `FindQuads`: ``` cv::threshold(mGrayPyramid[i], mBinaryPyramid[i], -1, 1, cv::THRESH_BINARY | cv::THRESH_OTSU); ```

Returning confidence seems interesting for advanced applications, indeed, but we also want to keep one "simple" API (aka [detect tags in 2 obvious lines](https://github.com/chili-epfl/chilitags/blob/master/samples/detection/detect-from-file.cpp#L40)). So, either we add a `findWithConfidence`...

@ayberkozgur : thanks for the review. I've updated the PR with the _vertical_ FOV instead of the horizontal one.

@qbonnard I almost agree with you... the OpenGL [tools](https://www.opengl.org/sdk/docs/man2/xhtml/gluPerspective.xml) [I](http://docs.unity3d.com/ScriptReference/Camera-fieldOfView.html) [know](https://www.opengl.org/sdk/libs/GLUS/docs/glus__matrix__viewprojection_8h.html#a77ae14512d9e18027210386fc3f051cb) all take the vertical FOV to setup the camera. So it may be convenient (especially since the FOV can...

@qbonnard Ok, replaced 5 lines with a single line + removed the SOLVE_ITERATIVE that actually slipped there.