Question about input data format of "convert_smpl_to_star.py"

[Miles629]

Thanks for your amazing work!
I want to convert smpl vertices to star but failed.
I think I may have mistaken the format of the input data. The input data I use is smpl vertices of people_shapshot dataset, and the dim is (6890,3).
I run the "convert_smpl_to_star.py", and the output is:

/project/macaoyuan/STAR/convertors/losses.py:78: UserWarning: The Default optimization parameters (MAX_ITER_EDGES,MAX_ITER_VERTS) were tested on batch size 32 or smaller batches
  'The Default optimization parameters (MAX_ITER_EDGES,MAX_ITER_VERTS) were tested on batch size 32 or smaller batches')
Loading the SMPL Meshes and 
STAGE 1/2 - Fitting the Model on Edges Objective
Traceback (most recent call last):
  File "./convertors/convert_smpl_to_star.py", line 56, in <module>
    np_poses , np_betas , np_trans , star_verts = convert_smpl_2_star(smpl,**opt_parms)
  File "/project/macaoyuan/STAR/convertors/losses.py", line 98, in convert_smpl_2_star
    d = star(poses, betas, trans)
  File "/home/macaoyuan/.local/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "/home/macaoyuan/.conda/envs/neuralbody/lib/python3.7/site-packages/star/pytorch/star.py", line 139, in forward
    v = v + trans[:,None,:]
RuntimeError: CUDA out of memory. Tried to allocate 544.00 MiB (GPU 0; 15.78 GiB total capacity; 13.47 GiB already allocated; 194.00 MiB free; 14.25 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

I think there is something wrong with my input data. Can you tell me the usage of the convert python file?

[Suraj-Bhor]

Hey @Miles629, the input SMPL mesh should be a numpy array of shape (batch_size x 6890 x 3), where the batch_size can be kept as 1.