Camera intrinsic matrix for in-the-wild images

llcshappy / Monocular-3D-Human-Pose

Generalizaing Monocular 3D Human Pose Estimation in the Wild

146 stars 16 forks source link

Camera intrinsic matrix for in-the-wild images #5

Open sdxass opened 5 years ago

sdxass commented 5 years ago

Hi,

First of all, thanks for open sourcing such a great piece of work. I have a question about the geometric search scheme for in-the-wild images:

In Section 3.4, the geometric search scheme need camera intrinsic matrix in order to perform the heuristic projection. So my question is how did you decide the camera intrinsic matrix for images in the wild?

Any response will be helpful. Thank you again for your wonderful work.

llcshappy commented 5 years ago

Thank you for your support of our work. Actually, we only use the camera intrinsic matrix of Human3.6M. The 3D key-points from Unity tool-box are reprojected by the camera intrinsic matrix of the Human3.6M to obtain the 2D key-points. In this way, we can use the same camera intrinsic matrix to perform the heuristic projection in the geometric search scheme.

sdxass commented 5 years ago

I understand it, thank you so much!