These scripts do everything for you (installation, model/data download and demo). If you run into errors, check out the colab example.
# install conda environment
./install_conda_env.sh
# download essentials and models
./fetch_data.sh
# download body models (SMPL-X, SMPL, SMIL). The script will ask for you username
# and password for the SMPL-X and SMPL website. If you don't have an account, please
# register under https://smpl-x.is.tue.mpg.de/ and https://smpl.is.tue.mpg.de/.
./fetch_bodymodels.sh
# Install BEV and ViTPose and convert body models to BEV format
./install_thirdparty.sh
# Run optimization with BUDDI on your own images
# We have some internet images in [this](./demo/data/FlickrCI3D_Signatures/demo/images_live) folder.
# The script will first run BEV and ViTPose and then start the optimization with BUDDI.
# To run the demo with OpenPose on top, please read the comments in demo.sh
./demo.sh
-
Install virtual environment with python packages
# Eventually update conda # conda update conda conda create -n hhcenv39 python=3.9 conda activate hhcenv39 conda install -c pytorch pytorch=1.9.1 torchvision cudatoolkit=10.2 conda install -c fvcore -c iopath -c conda-forge fvcore iopath conda install -c bottler nvidiacub conda install pytorch3d -c pytorch3d conda install -c conda-forge tensorboard pip install opencv-python smplx scipy scikit-image loguru omegaconf ipdb einops chumpy trimesh setuptools==58.2.0 conda run -n hhcenv39 --live-stream pip install 'git+https://github.com/facebookresearch/detectron2.git' pip install mmcv==1.3.9 timm pip install -v -e third-party/ViTPose/ pip install simple_romp==1.1.3 -
Eventually, you need to add environment variables in shell or to you .bashrc
YOUR_PROJECT_FOLDER=$(pwd) export PYTHONPATH="$YOUR_PROJECT_FOLDER/buddi" -
Install conda environment and download the supporting files via shell script
scripts/install_conda_env.sh
-
You can skip this step if you'll be using install.sh
-
Dowload buddi supporting files from here and extract them to
$REPO_ROOT/essentials├── $REPO_ROOT │ ├── essentials │ │ ├── buddi │ │ │ ├── buddi_unconditional.pt │ │ │ ├── buddi_unconditional.yaml │ │ │ ├── buddi_cond_bev.pt │ │ │ ├── buddi_cond_bev.yaml │ │ ├── priors │ │ │ ├── gmm_08.pkl │ │ ├── contact │ │ │ ├── flickrci3ds_r75_rid_to_smplx_vid.pkl │ │ ├── body_models │ │ │ ├── smil │ │ │ │ ├── smplx_kid_template.npy │ │ ├── body_model_utils │ │ │ ├── lowres_smplx.pkl │ │ │ ├── smplx_faces.pt │ │ │ ├── smpl_to_smplx.pkl │ │ │ ├── smplx_inner_mouth_bounds.pkl -
Link imar vision tools repo to
$REPO_ROOT/essentials:ln -s $REPO_ROOT/third-party/imar_vision_datasets_tools $REPO_ROOT/essentials/imar_vision_datasets_tools├── $REPO_ROOT │ ├── essentials │ │ ├── imar_vision_datasets_tools -
Bash script to download essentials and Link imar vision tools repo
scripts/fetch_data.sh
-
You need SMPL-X, SMPL, SMPL-A, SMIL. The SMIL-X Kid Template can be found on tha AGORA website.
-
Note that SMPL-X, SMIL, and SMPL-A underlie different licences (not MIT)
-
Download each model from the website and extract data to $SMPL_FOLDER. Then symlink $SMPL_FOLDER to essentials:
ln -s $SMPL_FOLDER $REPO_ROOT/essentials/body_models. -
Then run follow the instructions here to generate the body model files used BEV.
# generate SMPLA_NEUTRAL.pth and smil_packed_info.pth romp.prepare_smpl -source_dir=/path/to/smpl_model_data bev.prepare_smil -source_dir=/path/to/smpl_model_data -
The output should look like this:
├── $REPO_ROOT │ ├── essentials │ │ ├── body_models │ │ │ ├── smil │ │ │ | ├── smil_web.pkl │ │ │ │ ├── smil_packed_info.pth │ │ │ │ ├── smplx_kid_template.npy │ │ │ ├── smpl │ │ │ │ ├── SMPL_NEUTRAL.pkl │ │ │ ├── smplx │ │ │ │ ├── SMPLX_NEUTRAL.npz │ │ │ │ ├── SMPLX_NEUTRAL.pkl │ │ │ ├── smpla │ │ │ │ ├── SMPLA_NEUTRAL.pth
-
Bash script to download body models. You need an account on the websites of SMPL-X and SMPL. If you don't have an account, create one before you run
fetch_bodymodels.sh; the scirpt will ask for your credentials.scripts/fetch_bodymodels.sh