This library contains C++ code that is used to extract features from videos using optical Flow. Currently I have implemented two types of optical flow, Farneback and Lucas Kanade. The first is a dense optical flow method and the latter uses a more sparse representation. The idea of the of the C++ code was to attempt to capture the essense of the MATLAB code located in src/matlab.

Furthermore, R code is included to then classify the features. This code is located in src/r_code. The idea is to take the output of the C++ code and run it as the input to the R code to correctly classify the motions in the video. Particularly, this is an attempt to classify typing vs. no typing and writing vs no writing in videos provided through AOLME (Advancing Out of School Learning in Mathematics and Engineering).

In order to build the C++, you must have the following requirements:

• cmake
• OpenCV >= 3.1

In order to run the scripts, you must have:

• MATLAB >= 2015b
• R
• Python 2.7

If you have trouble, refer to the .travis.yml file. All the necessary building tools are installed in that file.

If you would like to run the code, particularly the extract_features program, you can do so if you are running Ubuntu by installing the .deb file in the releases.