Skip to content


Folders and files

Last commit message
Last commit date

Latest commit



22 Commits

Repository files navigation

Human Dynamics from Monocular Video with Dynamic Camera Movements

Ri Yu, Hwangpil Park and Jehee Lee

Seoul National University

ACM Transactions on Graphics, Volume 40, Number 6, Article 208. (SIGGRAPH Asia 2021)

Teaser Image


We propose a new method that reconstructs 3D human motion from in-the wild video by making full use of prior knowledge on the laws of physics. Previous studies focus on reconstructing joint angles and positions in the body local coordinate frame. Body translations and rotations in the global reference frame are partially reconstructed only when the video has a static camera view. We are interested in overcoming this static view limitation to deal with dynamic view videos. The camera may pan, tilt, and zoom to track the moving subject. Since we do not assume any limitations on camera movements, body translations and rotations from the video do not correspond to absolute positions in the reference frame. The key technical challenge is inferring body translations and rotations from a sequence of 3D full-body poses, assuming the absence of root motion. This inference is possible because human motion obeys the law of physics. Our reconstruction algorithm produces a control policy that simulates 3D human motion imitating the one in the video. Our algorithm is particularly useful for reconstructing highly dynamic movements, such as sports, dance, gymnastics, and parkour actions.


  • Ubuntu (tested on 22.04 LTS and 18.04 LTS)

  • Python 3 (tested on version 3.6+)

  • Dart (modified version, see below)

  • Fltk



sudo apt install git cmake wget build-essential freeglut3-dev libeigen3-dev libassimp-dev libccd-dev libfcl-dev libboost-regex-dev libboost-system-dev libopenscenegraph-dev coinor-libipopt-dev libbullet-dev libode-dev liboctomap-dev libflann-dev libtinyxml2-dev liburdfdom-dev doxygen libxi-dev libxmu-dev liblz4-dev

# Ubuntu 18.04
sudo apt install libnlopt-dev

# Ubuntu 22.04
sudo apt install libnlopt-cxx-dev libboost-filesystem-dev

git clone
cd dart-ltspd
mkdir build
cd build
cmake ..
make -j4
sudo make install


# Ubuntu 18.04
sudo apt install swig

# Ubuntu 22.04
sudo apt install swig3.0
sudo ln -s /usr/bin/swig3.0 /usr/bin/swig

sudo apt install python3-venv python3-dev

after making a virtual environment (venv) in the movingcam repo,

source venv/bin/activate
git clone
cd pydart2
pip install -U pip
pip install wheel

# Python 3.6
pip install numpy
pip install pyopengl==3.1.0
pip install pyopengl-accelerate==3.1.0

# Python 3.7 or above
pip install numpy==1.23
pip install pyopengl pyopengl-accelerate

python build
python install

Fltk and Pyfltk

sudo apt install libfltk1.3-dev

Download pyfltk

cd ~/Downloads
tar xzf pyFltk-
cd pyFltk-
python build
python install


sudo apt install libgle3-dev

# Python 3.6
pip install torch
pip install cvxopt
pip install scipy
pip install gym
pip install tensorboard==2.0.0
pip install setuptools==59.5.0
pip install pillow

# Python 3.7 or above
pip install torch==1.13.1 cvxopt scipy gym tensorboard pillow

Dump frames

sudo apt install ffmpeg ffplay
cd PyCommon/modules/GUI
python build
find ./ -name "csDump*.so" | xargs -I{} cp {} ./

Steps for running with a custom video

  1. Run OpenPose to get 2D poses (.json).

  2. Run "foot_contact/" to get contact information ([motion_name]_contact_info.npy) from step 1.

  3. Run VIBE to get 3D poses (.pkl).

  4. Run "ik/" to get the IK result ([motion_name].skmo) from step 3.

  5. Define the environment (interacting objects) in the "control/[motion_name]/[motion_name]". (You can search for related code lines by the word "box".)

  6. Run PPO "control/hurdle/[motion_name]".

Run examples

source venv/bin/activate
export LD_LIBRARY_PATH=/usr/local/lib
cd control/parkour1


    author = {Yu, Ri and Park, Hwangpil and Lee, Jehee},
    title = {Human Dynamics from Monocular Video with Dynamic Camera Movements},
    journal = {ACM Trans. Graph.},
    volume = {40},
    number = {6},
    year = {2021},
    articleno = {208}


No description, website, or topics provided.






No releases published


No packages published