I noticed SMPLicit uses a atart_posed SMPL to compute points distance for pants and skirts, but uses unposed SMPL for upper clothes. Is there any reason to do so?

Hello, thanks for your great work. I run all the examples and got expected results. And is it possible to load an existing SMPL obj file and do the similar clothes change to it? Thanks.

Hello,

Thank you for the code and the project, it looks awesome.

Could you please help with an issue I face when running python example.py:

FileNotFoundError: [Errno 2] No such file or directory: '/home/enric/cvpr21/SMPLicit/clusters/indexs_clusters_tshirt_smpl.npy'

Looks that one needs some additional files to run the demo. Thank you in advance!

Hi!

Thanks for your amazing work. I am trying to display the dressed-mesh results with various pose and shape parameters instead of the T-pose. It works fine for the upper clothes, hair, and shoes. However, for pants and skirts, it seems that they are upside down (second row in the Figure below). I don't know how to solve it.

I just change the following code in example.py, pants_Z = np.zeros(18) Zs = [pants_Z] meshes = SMPLicit_layer.reconstruct(model_ids=[1], Zs=Zs,pose=loaded_pose,beta=loaded_shape)

Thanks!

Thanks for the awesome work. It works well in example.py. However, I meet a problem when using "python fit_SMPLicit.py".

Traceback (most recent call last):
  File "fit_SMPLicit.py", line 10, in <module>
    from utils import projection
  File "/home/xxx/workspace/VG/SMPLicit/fit_SMPLicit/utils/projection.py", line 3, in <module>
    import pyrender
  File "/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/site-packages/pyrender/__init__.py", line 10, in <module>
    from .scene import Scene
  File "/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/site-packages/pyrender/scene.py", line 7, in <module>
    import networkx as nx
  File "/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/site-packages/networkx-3.0b1-py3.7.egg/networkx/__init__.py", line 21, in <module>
    from networkx import classes
  File "/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/site-packages/networkx-3.0b1-py3.7.egg/networkx/classes/__init__.py", line 1, in <module>
    from .graph import Graph
  File "/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/site-packages/networkx-3.0b1-py3.7.egg/networkx/classes/graph.py", line 11, in <module>
    from functools import cached_property
ImportError: cannot import name 'cached_property' from 'functools' (/home/xxx/anaconda3/envs/SMPLicit/lib/python3.7/functools.py)

Env: python3.7, kaolinv0.1.0, cuda10.2, torch1.10.1, SMPLicit0.0.1

Someone mention that it should use pthon3.8, however it conflicts with kaolinv0.1. May someone help me?

hi, thanks for your work. In section 4.2 you said see the Supp. Mat. for more details, but there is no supplement in arxiv version. So can I have the supplements or the details about training? thanks!

I tried git clone https://github.com/ecorona/smplicit, but it was a mistake "fatal: The repository "https://github.com/ecorona/SMPLicit/" not found" Why is that？

Hi, thanks for the great work! I ran the demo of fitting the input image and got good results. However, I currently have the task of fitting point clouds. Would you consider releasing the source script for fitting the scan? Or could you offer some suggestions for modifying the script of fitting the image to fitting the scan? Thanks a lot.

hello, i don't know what "topleft " and "scale_ratio" are? topleft = smpl_prediction[index_fitting]['bbox_top_left'] scale_ratio = smpl_prediction[index_fitting]['bbox_scale_ratio']

i only get the "bboxes" which value is [1034.61365568 543.89215647 591.51202979 591.51202979], so how can i get "topleft " and "scale_ratio" ?

Hi, using a MacBook Pro 16" 2019, and I received that AssertionErrror when running example.py

I have already used torch.load with map_location=torch.device('cpu') but still get that error message.

What should I do?

I am working on a problem which requires animating finger movements and facial expressions. Hence, I am using PIXIE, which uses the SMPL-X model. However, PIXIE works on the vanilla, undressed SMPL model. I was wondering what the easiest way is to add the clothing feature to PIXIE (i.e., I want PIXIE's output on a clothed SMPL model). It would suffice for the clothing/skin color to be a fixed value i.e., it doesn't have to be derived from the input image.

Hi Enric,

Thank you for the great work! For the model training, could you please share the code for obtaining occlusion UV maps as the model inputs, or briefly describe how to obtain occlusion UV maps on custom data? Thanks!