search

Jeff-sjtu / HybrIK

Official code of "HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation", CVPR 2021
MIT License
1.16k stars 142 forks source link

About the outputs #189

Open xyIsHere opened 10 months ago

xyIsHere commented 10 months ago

Dear author,

I successfully run the code and get the outputs of my input video. Then I printed all the shape of the output as below. image In my project, I used the VIBE to compute the shape and pose. But I think the results is not perfect. Now I want to switch to your method for shape and pose estimation. I'm wondering if there is any difference between the shape and pose results predicted by HybrIK and VIBE.

Thanks a lot!

biansy000 commented 10 months ago

pred_betas means the predicted shape parameter, and pred_thetas means the predicted pose parameter. pred_thetas gives the rotation matrix of 24 joints used in SMPL (216 = 24 x 3 x 3).

xyIsHere commented 10 months ago

Thank you very much! By the way, I also need the weak perspective camera parameters in cropped image space (s,tx,ty). How can I construct this output? And what is the difference between transl and transl_camsys?

biansy000 commented 10 months ago

In the output, transl means (tx, ty, tz) and we use focal length = 1000

xyIsHere commented 10 months ago

In the output, transl means (tx, ty, tz) and we use focal length = 1000

Thank you so much!