Thanks for your creative work! I encountered a question that the retargeted model acted not like source CMU motion data after retargeting. I did not modify any repo code, and visualized the H1 t-pose model. It looks fine, however, when I used "retarget_motion_h1.py" to generate the targeted motion data, H1 moved much more differently than the source data. I used "retarget_cmu_to_h1.json", "fbx_importer.py", and "generate_h1_humanoid_tpose.py" to generate needed files.
this is a cmu t-pose model:
this is a targeted H1 model:
Thanks for your creative work! I encountered a question that the retargeted model acted not like source CMU motion data after retargeting. I did not modify any repo code, and visualized the H1 t-pose model. It looks fine, however, when I used "retarget_motion_h1.py" to generate the targeted motion data, H1 moved much more differently than the source data. I used "retarget_cmu_to_h1.json", "fbx_importer.py", and "generate_h1_humanoid_tpose.py" to generate needed files.
this is a cmu t-pose model: this is a targeted H1 model:
this is the retargeted H1 video: https://github.com/user-attachments/assets/c840777b-3792-447f-8cef-c69902790cd5
I am not clear which step is missed or wrong, looking forward to your reply. Many thanks!