opencv_traincascade unhandeled exception
I am training a cascade classifier using opencv. I have 5 positive and 5 negative samples (this is for test on one image, if it will work i will use 500 good and 500 bag images). Their dimensions are 20 * 20 pixels.
The code that I used to make the samples was:
D:\learn\opencv_createsamples.exe -info D:\learn\Good.dat -vec D:\learn\samples.vec -w 20 -h 20
this is the traning code:
D:\learn\opencv_traincascade.exe -data D:\learn\haarcascade -vec D:\learn\samples.vec -bg D:\learn\Bad.dat -numStages 16 -minhitrate 0.8 -maxFalseAlarmRate 0.2 -numPos 5 -numNeg 5 -w 20 -h 20 -mode ALL -precalcValBufSize 512 -precalcIdxBufSize 512
Then this application crashes and i see message with error on stage 1:
Which of the arguments are wrong?
I don't know if it is the problem but there is a typo error with
-minhitrate 0.8
. You can see it as the program took the default value for-minHitRate
.Wait a minute, are you using EgmuCV? Then probably a bug with the wrapper, but in any case you should ask in the proper forum (maybe http://www.emgu.com/forum/viewforum.p...). In this forum we don't give support to OpenCV wrappers (see FAQ)
Also
5 positive and 5 negative samples
for testing purposes --> won't work ... you will run into problems created by that small amount... use the search button on this forum ontraincascade parameters
and tons of topics will pop up that discuss a good use of parameters.