Pose 3D output is centered to origin.

DenisTome / Lifting-from-the-Deep-release

Implementation of "Lifting from the Deep: Convolutional 3D Pose Estimation from a Single Image"

https://denistome.github.io/papers/lifting-from-the-deep

GNU General Public License v3.0

450 stars 132 forks source link

Pose 3D output is centered to origin. #25

Closed KelSolaar closed 6 years ago

KelSolaar commented 6 years ago

Hi,

I'm testing the project on an image sequence, and while I manage to successfully get poses for the sequence, they are always centered to origin. I tried to initialise a PoseEstimator for the whole sequence or one per frame without much differences. Am I missing something obvious?

Cheers,

Thomas

DenisTome commented 6 years ago

Hi,

this is due to a normalization that it's applied in the code.

Guo-king commented 5 years ago

I have tested the project on a video，and I found that the 3D outputs centered to origin. Is there some methods to save the problem. Or shall i remove some code to avoid the normalization because I want to reserve the change of skeleton center. @KelSolaar @DenisTome THANK YOU!