You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on May 9, 2023. It is now read-only.
Looking through the code, there appears at first to be support for SMPL, however the code does not seem to account for the parameters of that model. For example, SMPL models are issued in PKL files, however your AMASS data loader expects NPZ (for newer version such as SMPL-H / SMPL-X). I modified the code to load the PKL, but further along in the loading process, fairmotion is unable to parse the SMPL model:
File "/home/visionarymind/eclipse-workspace/fairmotion/load_amass.py", line 19, in <module>
bm = amass.load_body_model(bm_path,num_betas,model_type)
File "/home/visionarymind/anaconda3/envs/fairmotion/lib/python3.7/site-packages/fairmotion/data/amass.py", line 130, in load_body_model
num_betas=num_betas,
File "/home/visionarymind/anaconda3/envs/fairmotion/lib/python3.7/site-packages/human_body_prior/body_model/body_model.py", line 105, in __init__
self.comp_register('shapedirs', torch.tensor(shapedirs, dtype=dtype), persistent=persistant_buffer)
ValueError: too many dimensions 'Select'
Before digging further, I wanted to see if you have tested SMPL data loads and what, if anything, could be done with this code to make it compatible with its models.