🌟 smpl_ros

A ROS 2 package for integrating SMPL with real-time Rviz visualization

---

🔍 Overview

smpl_ros brings the SMPL human body model into the ROS 2 ecosystem with seamless Rviz visualization support.
It’s designed to make working with 3D human meshes in robotics and visualization pipelines simple, fast, and flexible.

---

📦 Dependencies

To get started, make sure you have the following installed:

• ROS 2 (tested with Humble)
• Open3D C++ library
  • Used for point cloud loading and downsampling
  • Alternatively, you can use PCL with minor code changes
• CUDA (optional) – Enables GPU acceleration
• torchure_smplx – C++ SMPL Torch implementation

---

⚙️ Installation

1. Convert your SMPL model in .npz format. You can download the .pkl from the SMPL website (registration required) and convert it using the provided script
2. Export required paths to Open3D and torchure_smplx CMake files

   export Open3D_DIR=~/path_to_your_open3d_build/lib/cmake/Open3D
   export torchure_smplx_DIR=~/path_to_your_torchure_smplx_build
   

3. (Optional) Configure CUDA

   CUDA_VERSION=<your_cuda_version>
   export PATH="/usr/local/cuda-${CUDA_VERSION}/bin:$PATH"
   export LD_LIBRARY_PATH="/usr/local/cuda-${CUDA_VERSION}/lib64:$LD_LIBRARY_PATH"
   

4. Clone the repository

   cd ~/your_ros2_ws/src
   git clone https://github.com/idra-lab/smpl_ros.git
   git clone https://github.com/idra-lab/smpl_msgs.git
   

5. Build the workspace

   cd ~/your_ros2_ws
   colcon build --symlink-install
   source install/setup.bash
   

🚀 Running the Node

A fitting procedure example is provided in fitting_smpl_ros.cpp, it can be run with:

ros2 launch smpl_ros smpl_ros_fitting.launch.py \
    model_path:=/path/to/SMPL_model.npz \
    config_path:=/path/to/config.json \
    point_cloud_path:=/path/to/point_cloud.ply

which will also open an Rviz window to visualize the process.