Installation steps:

SMPLify-x

Create new environment

conda create --name smplx0 python=3.8
conda activate smplx0

Clone the repository and run setup.py

git clone https://github.com/AdityaC19/smplify-x.git
python setup.py install

Install packages

pip install smplx
pip install pyrender
conda install -c conda-forge trimesh
pip install -r requirements.txt

Model Loading

Link for SMPL: https://smpl.is.tue.mpg.de/

Download the model from the link and save in a directory with the following structure

models
|__ smpl
      |__ SMPL_FEMALE.pkl
      |__ SMPL_MALE.pkl
      |__ SMPL_NEUTRAL.pkl

Fitting

After installing the smpl package and downloading the model parameters you should be able to run the main.py script to visualize the results. For this step you have to install the pyrender and trimesh packages.

Run the following command to execute the code:

python smplifyx/main.py \
      --config cfg_files/fit_smpl.yaml \
      --data_folder DATA_FOLDER
      --output_folder SMPL_OUTPUT \
      --visualize=True/False \
      --model_folder models \
      --vposer_ckpt vposer_v1_0 --part_segm_fn smplx_parts_segm.pkl

where the DATA_FOLDER should contain two subfolders, images, where the images are located, and keypoints, where the OpenPose output should be stored.

Note: 2D Keypoints for input image can be obtained from Openpose. Detailed instructions for installation are mentioned in Openpose repository forked above

Skeleton Based Action Recognition