Depth Based Object Tracking Library (dbot)

The core of this library are two probabilistic tracker

• A Non-Parametric Tracker Based on Rao-Blackwellized Coordinate Descent Particle Filter

[M. Wuthrich, P. Pastor, M. Kalakrishnan, J. Bohg, and S. Schaal. Probabilistic Object Tracking using a Range Camera IEEE Intl Conf on Intelligent Robots and Systems, 2013]

http://arxiv.org/abs/1505.00241

This tracker can run on a pure CPU system or optionally run on GPU using CUDA 6.5 or later

• A Parametric Tracker Based on Robust Multi-Sensor Gaussian Filter Tracker

  [J. Issac, M. Wuthrich, C. Garcia Cifuentes, J. Bohg, S. Trimpe, S. Schaal Depth-Based Object Tracking Using a Robust Gaussian Filter IEEE Intl Conf on Robotics and Automation, 2016]

  http://arxiv.org/abs/1602.06157

All trackers require mesh models of the tracked objects in Wavefront (.obj) format.

Requirements

• Ubuntu 14.04
• C++11 Compiler (gcc-4.7 or later)
• CUDA 6.5 or later (optional)

Dependecies

• Filtering library (fl)
• Eigen 3.2.1 or later

Compiling

The cmake package uses Catkin. If you have installed ROS groovy or later, then you most likely have catkin installed already.

 $ cd $HOME
 $ mkdir -p projects/tracking/src  
 $ cd projects/tracking/src
 $ git clone git@github.com:filtering-library/fl.git
 $ git clone git@github.com:bayesian-object-tracking/dbot.git
 $ cd ..
 $ catkin_make -DCMAKE_BUILD_TYPE=Release -DDBOT_BUILD_GPU=On

If no CUDA enabled device is available, you can deactivate the GPU implementation via

 $ catkin_make -DCMAKE_BUILD_TYPE=Release -DDBOT_BUILD_GPU=Off

How to use dbot

Checkout the ros nodes of each tracker in dbot_ros package for exact usage of the filters.