search

DirtyHarryLYL / DJ-RN

As a part of HAKE project (HAKE-3D). Code for our CVPR2020 paper "Detailed 2D-3D Joint Representation for Human-Object Interaction".
Apache License 2.0
100 stars 13 forks source link

What is shape and pose parameter? #72

Closed mansooreh1 closed 3 years ago

mansooreh1 commented 3 years ago

In your paper, you told image What is image and image What do the pose and shape parameters mean?

walnut-REE commented 3 years ago

In your paper, you told image What is image and image What do the pose and shape parameters mean?

If I understand correctly it's described in the original SMPL paper?

mansooreh1 commented 3 years ago

I'm a little confused. Can you explain? Is image the point coordinates of joints in 2D and image the point coordinate of joints in 3D?

DirtyHarryLYL commented 3 years ago

2D keypoints are coordinates from Openpose. The 3D pose is recorded as relative rotation angles from Smplify-x, which can be easily converted to XYZ. Please refer to the Smplify-x for more.

mansooreh1 commented 3 years ago

Thanks .I fully understood. Just what are image? In smplify-X told these two parameter are shape parameter(betas) and facial expression parameter(expression). Can you explain bit what that means?

DirtyHarryLYL commented 3 years ago

Compared to SMPlify, SMPLify-X not only encode the body shape but also the hand-face shape. According to their setting, the 3D points on hand and face are also regressed, based on the Openpose 2D hand-face keypoints. Detailed illustration can be found in the Smplify-x output document.

mansooreh1 commented 3 years ago

Thanks