Please send your comments or requests to zhou.ji [at] yahoo.com
The package includes two pieces of work with key part shared.
1. A C implementation of V-Detector: four files vdetector.c vdetector_main.c, statutil.c, vdetector.h
2. A demonstration program written in C++ and Qt: three additional files vdmovie.cpp, vdmovie_main.cpp, vdmovie.h
- C implementation can be compiles with compile.bat. The resulted executive is vd.exe.
- The simple demonstration uses the C implementation (excluding vdetector_main.c). It is a complete Qt qmake project and can be compiled with command 'make'. The resulted executive in under the 'release' folder. I call this demonstration program 'vdmovie'.
Features or lack of it:
- Vdmovie can save the visulization image by one button click plus entering a file name.
- Vdmovie doesn't implement any thread mechanism. So when the new "movie" starts, you cannot do anything other watching the animation. If GUI is interrupted, it will not update until the entire "movie" finishes and the finished detector set is displayed properly.
Images
These images were generated directly by vdmovie. They are the main content of the program's display as well.
In this image above, Euclidean distance is used in matching rule. The self region is a pentagram ('star').
The green dots are the training data used to generate the detector set. The yellow circles are the detectors generated.
The threshold is set low, so the detectors are rather aggresive and 'penetrate' into the self region.
Above is a case when the Manhattan distance (L-1 distance) is used.
Above is a case of 'ring'-shaped self region.
© 2007 Zhou Ji