search

Advocate99 / DiffGesture

[CVPR'2023] Taming Diffusion Models for Audio-Driven Co-Speech Gesture Generation
GNU General Public License v3.0
234 stars 16 forks source link

Real Pose Generate #19

Closed qingyuanxingsi closed 10 months ago

qingyuanxingsi commented 10 months ago

Thanks for your great work, how can I generate real poses corresponding to the openpose format? I found that output_poses generated by inference script are quite small numbers

Advocate99 commented 10 months ago

Hi, in our generation space, the poses are in the format of direction vectors. So, if we know the length of the controlling bones, we can convert the vectors to the positions of the bones. You can check the code scripts/utils/data_utils_expressive.py .