CascadeSVMs is an open source C++ implementation of the CascadeSVMs algorithm (bounded with LIBLINEAR) for highly imbalanced large-scale data (i.e., negative samples >> positive samples) learning.
To compile the code, you need to have the OpenCV library (tested with OpenCV-2.4.6).
The interface is designed for TRECVID Surveillance Event Detection. It can be modified according to your tasks.
./CascadeSVMs -c train.control -p train
./CascadeSVMs -c test.control -p test
The configuration and parameter settings are embeded in the training and testing control files.
If you find any bug or develop some extensions, please feel free to drop me a line.
Please cite our paper if you use the code:
X. Yang, C. Yi, L. Cao, and Y. Tian. MediaCCNY at TRECVID 2012: Surveillance Event Detection. NIST TRECVID Workshop, 2012. [PDF]
Copyright (C) 2013 Xiaodong Yang
Distibution code version 1.0 - 05/09/2014. The code is for research purpose only.