Human-Robot-Animations

This repository contains a library of human motion capture data recorded with a Vicon tracker system.

Furthermore, it provides a processing pipeline for transforming the recorded data from .csv files to .bvh and further to .pkl files that can be used in mujoco simulations.

Clone this repo

[email protected]:JakobThumm/human-robot-animations.git

Installation

We work on Ubuntu 20.04. If you want to participate, you need to run this version

1. Install anaconda

2. Create an environment

   conda create -n human-motion python==3.8
   

3. Activate the environment

   conda activate human-motion 
   

4. Move to the transformation_pipeline subdirectory as it contains the transformation pipeline script

   cd transformation_pipeline/
   

5. Install the requirements

   pip install -r requirements.txt 
   

6. Run the pipeline to convert the collected csv data to a mujoco usable format

   python transformationPipeline.py --directory motion_capture/dataset/motions/{name of the folder} 
   

   use the following arguments:

   --postprocess (apply postprocessing)

   --toBVH (convert CSV to BVH)

   --toMujoco (convert BVH to mujoco usable format)

The pipeline creates an extra folder called mujoco for the mujoco usable files (.pkl)

Please make sure you have a file tPose*.csv present in the folder containing the other .csv files you would like to convert. It should contain data of a human performing a t-Pose. We use only the first frame of this file in the transformation pipeline. Make sure all markers are visible in that frame!