(1) Reference:

Discriminative Subvolume Search for Efficient Action Detection, Junsong Yuan, Zicheng Liu, Ying Wu, 
IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2009), Miami, Florida, June 22-24, 2009

(2) Syntax:

SubVolumeSearch infile outfile numclasses width height penaltyValue classthreshold1 classthreshold2 ... classthresholdn

An example:

SubVolumeSearch.exe example.keyscore example.DetRes 3 160 120 -0.0003 15.0 15.0 15.0

In the above example, the infile is "example.keyscore", outfile is "example.DetRes", numclasses=3, width=160, height=120, penaltyValue = -0.0003, 
classthreshold1=15.0, classthreshold2=15.0, classthreshold1=15.0.

(3) the format of the infile "example.keyscore":

The first row: number of interest points
The second row: number of frames
Subsequently, each row corresponds to an interest point:

x y frameNumber arbitrary_value class1_score class2_score class3_score

Note that the fourth number on each row (<arbitrary_value>) is not used by the SubVolumeSearch algorithm. You can put any value here. <class1_score> is the score of
the interest point with respect to class#1.

(4) the format of the outfile "example.DetRes":

Each row corresponds to a detected subvolume. The foramt is x width y height startFrame numFrames totalScore classLabel










