Skip to content


Repository files navigation

PoseTrack: Joint Multi-Person Pose Estimation and Tracking

The repo provides the source code for our CVPR'17 paper for joint mult-person pose estimation and tracking.

Umar Iqbal, Anton Milan, Juergen Gall
PoseTrack: Joint Multi-Person Pose Estimation and Tracking
IEEE Conference on Computer Vision and Pattern Recongnition (CVPR) 2017.
Project Page

The code is tested on Ubuntu 16.04 (64bit) with MATLAB (2016a).


Installation Instructions
  1. Clone repository

    $ git clone --recursive
  2. Build and download dependencies of DeeperCut

    1. Build Caffe and its MATLAB interface after configuring Makefile.config

      $ cd deepcut/external/caffe
      $ make -j 12 all matcaffe
    2. Build liblinear, specify the path to the MATLAB installation

      $ cd deepcut/external/liblinear-1.94/matlab
      $ CC=gcc CXX=g++ MATLABDIR=PATH_TO_MATLAB make
    3. Download models

      $ cd deepcut/data
      $ ./
  3. Build PoseTrack solver

    $ cd posetrack/solver
    $ cmake . -DGUROBI_ROOT_DIR=/path/to/gurobi605/linux64 -DGUROBI_VERSION=60
    $ make pt-solver-callback
  4. Obtain Gurobi license and place the license at desired location and modify the p.gurobi_license_file in posetrack/lib/pt_exp_params.m to point to the license file location.

  5. Download PoseTrack models

    $ cd data
    $ ./

Run Demo

% start matlab
>> startup
>> demo_posetrack

One video for demo purposes is already provided. However, if you want to run the code over all videos, please download Multi-Person PoseTrack dataset and place the videos under 'data/bonn-multiperson-posetrack/'.


	author = {Umar Iqbal and Anton Milan and Juergen Gall},
	title = {PoseTrack: Joint Multi-Person Pose Estimation and Tracking},
	booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
	year = {2017},
	url = {}


The authors are thankful to Chau Minh Triet, Andreas Doering, and Zain Umer Javaid for the help with annotating the dataset. The work has been financially supported by the DFG project GA 1927/5-1 (DFG Research Unit FOR 2535 Anticipating Human Behavior) and the ERC Starting Grant ARCA (677650).
Thanks to Eldar Insafutdinov for releasing the code for DeeperCut.


No description, website, or topics provided.






No releases published