traincascade: how to set bType to RAB?

asked 2018-01-13 09:30:42 -0500

BuzzSid gravatar image

Hello everybody, i have serious problem.

I have to build a classifier and I came across this paper (link text).

It recommends building a classifier with a Btype parameter set to RAB (page 22 of paper).

At the command prompt I gave this command: opencv_traincascade.exe -data cascade -vec vector.vec -bg bg.txt -numPos 231 -numNeg 463 -numStages 40 -featureType LBP –bt RAB -w 120 -h 40 -minHitRate 0.999 -maxWeakCount 100

once the command is launched, unfortunately the Btype value is left on the default GAB value, not modifying it. How can I change it?

What did I do wrong? How should I set the bType parameter to RAB? is it possible to set it if you build two LBP or HOG classifiers?

edit retag flag offensive close merge delete


It might be that different boosting options were only supported for HAAR, but I find it hard to find evidence of that in the sourcecode...

StevenPuttemans gravatar imageStevenPuttemans ( 2018-01-15 04:29:21 -0500 )edit

it's something that I thought too, I think you're right. As soon as I finish this training I try to set a haar classifier by setting the parameter RAB

BuzzSid gravatar imageBuzzSid ( 2018-01-15 09:14:28 -0500 )edit