Alexis David Jacq

Thankyou for your interest ! We use [ROS TF](http://wiki.ros.org/tf) library to do it. You can see in **src/ros_head_pose_estimator.cpp** the following lines : ``` auto poses = estimator.poses(); [...] tf::Quaternion qrot;...

rvec is a rotation vector (output of solvePNP), we use an OPENCV function "Rodigues" to get the 3x3 rotation matrix given this vector. Then, you can use OPENCV's [decomposeProjectionMatrix](http://docs.opencv.org/2.4/modules/calib3d/doc/camera_calibration_and_3d_reconstruction.html#decomposeprojectionmatrix) to...

Actually, the points we fund on wikipedia (position of the sellion etc...) are just used to compute the position in space of a head that is already detected. In order...

Ah ok ! So you want to say "with hands" to the computer that the subject is an adult woman or a child, and then the computer uses specific points...

it's a multidimensional normal distribution with a spherical covariance: you have `Nd_Sigma = 1d_sigma^2 * Nd_Identity`. so `|Nd_Sigma| = 1d_sigma^(2*k)` then, `k/2 * log(2 * pi * e) + 1/2...

Thanks for the notification, I will try with this normalization. Can-I ask you with which one did you compare?

Thanks a lot! I did not see this typo!

Hello, thanks for the issue. I am not able to run so many epochs with my personal computer so I can't reproduce it. However, as the loss is the log...

I did not try these data. It's a harder task since if you miss a segment or if an angle is not right, you miss the letter, but I think...

> but I found that the KL loss in your code was increasing Interesting, did you observe this after a high number of epochs, or directly at the beginning? (in...