Skip to content

Compressed-Domain Video Object Tracking using Markov Random Fields with Graph Cuts Optimization

License

Notifications You must be signed in to change notification settings

serhan-gul/hhi-stmrftracking

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Compressed-Domain Video Object Tracking using Markov Random Fields with Graph Cuts Optimization

Research group at Fraunhofer HHI: https://www.hhi.fraunhofer.de/en/departments/vca/research-groups/multimedia-communications.html

The application performs video object tracking with compressed-domain processing of H.264/AVC video bitstreams applying Markov Random Fields and Graph Cuts methods.


Compiling

Compile FFmpeg decoder for motion vector extraction
  1. Clone repository into libs folder
cd libs
git clone https://github.com/bombardellif/FFmpeg.git
cd FFmpeg
mkdir bin
  1. Install dependencies and compile (assuming Linux Ubuntu)
sudo apt install autoconf \
    automake \
    build-essential \
    cmake \
    libass-dev \
    libfreetype6-dev \
    libtheora-dev \
    libtool \
    libvorbis-dev \
    mercurial \
    pkg-config \
    texinfo \
    zlib1g-dev \
    libx264-dev \
    yasm
PATH="./bin:$PATH" PKG_CONFIG_PATH="./ffmpeg_build/lib/pkgconfig" ./configure
--prefix=./ffmpeg_build --bindir=./bin --extra-cflags=-I./ffmpeg_build/include
--extra-ldflags=-L./ffmpeg_build/lib --enable-gpl --enable-libx264 --enablenonfree --enable-shared
PATH="./bin:$PATH" make -j4 && make install
Install application dependencies
  1. Install Python and dependencies
sudo apt install python3-pip python3-dev python3-tk g++ gfortran liblapack-dev \
    liblapacke-dev libatlas-dev libopenblas-dev python3-cffi-backend \
    python3-cairo-dev libffi-dev python-opencv libopencv-dev
cd ../../
sudo pip3 install -r requirements.txt
  1. Compile the sub modules
cd utils
python3 setup.py build_ext --inplace
cd ../mincut
python3 setup.py build_ext --inplace
cd ../decoder
python3 setup.py build_ext --inplace
  1. Run the application (help output)
cd ..
LD_LIBRARY_PATH=libs/FFmpeg/ffmpeg_build/lib python3 -m hhi_stmrftracking.main -h

Datasets

Download the pictures and ground truth at http://www.votchallenge.net/vot2016/dataset.html.

Copy the shell script docs/encode-all.sh inside the downloaded folder and execute it to encode the pictures in H.264/AVC.

Download the ST-MRF code and dataset at https://www.researchgate.net/publication/258938622_ST-MRF_tracking. Videos are already encoded.


Example

With the dataset copied to the data folder. The last parameter is the ground truth for the first picture of the video sequence.

LD_LIBRARY_PATH=libs/FFmpeg/ffmpeg_build/lib python3 -m hhi_stmrftracking.main -d data/in/gymnastics4.264 -e data/gt-gymnastics4/ data/in/gymnastics4.264 data/in/gt-gymnastics4.png

The expected output can be found in docs/example_gymnastics4.264.eval.mp4.

About

Compressed-Domain Video Object Tracking using Markov Random Fields with Graph Cuts Optimization

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 57.2%
  • C++ 36.7%
  • C 5.3%
  • Other 0.8%