search

DeepMotionEditing / deep-motion-editing

An end-to-end library for editing and rendering motion of 3D characters with deep learning [SIGGRAPH 2020]
BSD 2-Clause "Simplified" License
1.58k stars 256 forks source link

In retarget , how can we get mean & var .npy ? #37

Closed xinzi2018 closed 4 years ago

PeizhuoLi commented 4 years ago

We are planning to release related details in 1~2 weeks. More details in #31. Thank you!

xinzi2018 commented 4 years ago

thanks a lot !

ANYMS-A commented 4 years ago

If you print out the shape of mean or var, you will find that the shape of them are like[(simple_joints_number-1)4+3, 1], because they're calculated by using the rotation of all edges("edges" means all simplified skeleton joints without the ROOT joint) and the ROOT joint's position w.r.t all bvh files' frames. So, a possible way to calculate the mean & var by yourself, is to read the edges' rotation and root position of every frame of every bvh file, then calculate the mean and var, then reshape into a numpy array with shape:[(simple_joints_number-1)4+3, 1]

PeizhuoLi commented 4 years ago

You can get mean and var use retargeting/datasets/preprocess.py. For more details please refer to dataset section in instructions for training from scratch.