search

ChrisWu1997 / 2D-Motion-Retargeting

PyTorch implementation for our paper Learning Character-Agnostic Motion for Motion Retargeting in 2D, SIGGRAPH 2019
https://motionretargeting2d.github.io
MIT License
440 stars 86 forks source link

About mean pose #8

Closed Zhangkunyao closed 5 years ago

Zhangkunyao commented 5 years ago

Hi, I noticed code written in predict.py. ''' mean_pose, std_pose = get_meanpose(config) ... preprocess_motion2d(input3, mean_pose, std_pose) ''' I was wondering,when I test my own video,should I also get those data first?

ChrisWu1997 commented 5 years ago

Yes. mean_pose and std_pose are always needed since we normalize the input by subtracting the mean and dividing the std:

(motion - mean_pose[:, :, np.newaxis]) / std_pose[:, :, np.newaxis]

But note the mean_pose and std_pose are from the training data, which is already included in the folder mixamo_data. So basically you don't need to do anything else except supplying the input motion.

QUTGXX commented 1 year ago

@ChrisWu1997 If I want to train the network based on my own dataset and using different nums of key points, how to get the mean pose and std_pose?