Jeff-sjtu / HybrIK

Official code of "HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimation", CVPR 2021
MIT License
1.17k stars 142 forks source link

Docker image #127

Open caihaunqai opened 1 year ago

caihaunqai commented 1 year ago

Hello, I have trouble preparing the environment for HybrIK. mainly from Pytorch3d. Could you please share a Docker image for running HybrIK? Thank you.

caihaunqai commented 1 year ago

Although I have followed all steps you provided, I still get into trouble with Pytorch3D. Can you help me? Thanks.

Jeff-sjtu commented 1 year ago

Hi @caihaunqai, you can follow the steps provided in the official PyTorch3D repo:

conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub

Then install PyTorch3d via pip:

pip install git+ssh://git@github.com/facebookresearch/pytorch3d.git@stable
caihaunqai commented 1 year ago

Hi @caihaunqai, you can follow the steps provided in the official PyTorch3D repo:

conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub

Then install PyTorch3d via pip:

pip install git+ssh://git@github.com/facebookresearch/pytorch3d.git@stable

Thank you for your reply. I have followed the above official steps, but it still meets errors. I don't know how to solve it. Can you share a docker image?

caihaunqai commented 1 year ago

Hi @caihaunqai, you can follow the steps provided in the official PyTorch3D repo:

conda install -c fvcore -c iopath -c conda-forge fvcore iopath
conda install -c bottler nvidiacub

Then install PyTorch3d via pip:

pip install git+ssh://git@github.com/facebookresearch/pytorch3d.git@stable

Thank you for your reply. I have followed the above official steps, but it still meets errors. I don't know how to solve it. Can you share a docker image?

I have tested your HybrIK method on colab demo and I found it can get the best results on some videos, compared with PyMaf, ROMP, PARE, and so on. And it is my preferred method. Hope you can help me. Thanks again.

Jeff-sjtu commented 1 year ago

Sorry, I haven't built the docker image for HybrIK yet. Could you provide the error log when installing PyTorch3D?

caihaunqai commented 1 year ago

The error I met is "RuntimeError: Not compiled with GPU support.". And I have tried lots of methods to solve it, but it didn't work.

Sorry, I haven't built the docker image for HybrIK yet. Could you provide the error log when installing PyTorch3D?

Jeff-sjtu commented 1 year ago

Which operating systems are you using? Linux or Windows? Besides, did you install the GPU version of PyTorch correctly? PyTorch3D requires at least PyTorch 1.8.0.

caihaunqai commented 1 year ago

I use linux(ubuntu) operating systems. The Pytorch version is 1.9.0 and GPU version is 11.0. Pytorch3d's version is 0.7.1.

Which operating systems are you using? Linux or Windows? Besides, did you install the GPU version of PyTorch correctly? PyTorch3D requires at least PyTorch 1.8.0.

Jeff-sjtu commented 1 year ago

You can try:

pip install --upgrade pip
pip install scikit-image matplotlib imageio black isort cython
conda install pytorch3d -c pytorch3d-nightly
caihaunqai commented 1 year ago

You can try:

pip install --upgrade pip
pip install scikit-image matplotlib imageio black isort cython
conda install pytorch3d -c pytorch3d-nightly

I have tried it, but it can not be installed correctly. You can see it in the picture.

image
Jeff-sjtu commented 1 year ago

Based on the issues in PyTorch3D (#1162 and #1139, something strange is happening with the Pytorch/torchvision installation. You can re-install PyTorch or using a higher version of PyTorch.