AdaBoost Algorithm for Object Tracking

This a C++ implementation of the AdaBoost algorithm from A desicion-theoretic generalization of on-line learning and an application to boosting by Yoav Freund and Robert E. Schapire. The algorithm is used to track an object in a video sequence.

The task is to track Nemo in a sequence 32 frames. The first ten annotated frames are used for training and the remaining frames are used for testing.

Requirements

• CMake 2.6+
• OpenCV 2.4.x
• C++11

How to compile:

In Linux terminal :

$ cmake CMakeLists.txt

Then compile it with :

$ make

How to run:

In Linux terminal :

$ ./bin/tracking nemo/frames.train nemo/frames.test 50

After running the program you will see 22 frames with the bounding box around the object(Nemo).