fagg / bitplanes

Robust and real time template tracking

BitPlanes

A library for robust real time tracking.

If you find this work useful, please cite our work

@proceedings{alismail2016bitplanes,
  title={Robust Tracking in Low Light and Sudden Illumination Changes},
  author={{Alismail}, Hatem and {Browning}, Brett and {Lucey}, Simon},
  booktitle={Internal Conference on 3D Vision (3DV)},
  year={2016}
}

See here for additional details and data.

Compiling

mkdir build && cd build && cmake -DCMAKE_BUILD_TYPE=Release ../ && make -j3

Dependecies

• Eigen 3.0+
• OpenCV 3.0+

Using the library

Look into the directory test/ for examples of running the code.

First, initialize the tracker with AlgorithmParameters (see core/algorithm_parameters.h for docs)

The default values should ok, but might need tweaking.

  using namespace bp;

  AlgorithmParameters params;
  params.max_iterations = 50;
  params.verbose = true;
  params.function_tolerance = 1e-5;
  params.parameter_tolerance = 1e-4;

  // Create the tracker

  BitplanesTracker<Homography> tracker(params);

 // Initialize the template

   /* The image must be grayscale
    * ROI indicate the template location within the image
    */
  tracker.setTemplate(image, roi);


// Track new frames

  for(auto I : images)
    auto result = tracker.track(I);

see core/types.h for the Result structure, which contains the estimated Homography along with other useful information.

For version optimized for Visual Odometry see bpvo