This is a C++ implementation of the tracking algorithm. Tracking can be performed on video sequences.

• OpenCV
• Eigen
• Boost filesystem system components

This code has been tested using OpenCV 3.0.0, Boost 1.59.0 and Eigen 3.2.6.

Note: We notice that if you are using OpenCV 3.1.0 built with Homebrew, VideoCapture and waitKey crashes after a while on OS X 10.11 as listed here. The simplest solution is reverting OpenCV back to 3.0.0.

CMake is used for cross-platform compilation. For example on Unit-based systems run:

> mkdir build
> cd build
> cmake ..
> make

Note: make sure you compile the code in Release mode, as a Debug build will result in significantly slower performance.

After compilation, from the top level of the repository run:

> build/bin/PAWSS [config-file-path]

If no path is given the application will attempt to use ./config.txt.

Please see config.txt for configuration options.

For manually selecting region of interest (ROI), RIGHT click and drag on the image of window "choose roi".

The code is tested using the Online Tracking Benchmark (OTB) dataset.

This code uses the Struck algorithm provided by Sam Hare.