Cascade Classifier - Utilize less processor with TBB training.

asked 2017-06-02 08:26:26 -0500

Simão Martin gravatar image

Hello, i am using OpenCV cascade training with LBP (or HAAR), and use TBB for multi-threading. The computer that i am using have 64 cores, cool. However more people use the computer. My training use 100% of all able processors (that do not being used), and this disturb other programs that are running. I want use less processor for the training (5~10 processors). How do I do this? If possible. Excuse me my english :(.

Thank you people.

edit retag flag offensive close merge delete


Try to use setNumThreads(8)

LBerger gravatar imageLBerger ( 2017-06-02 08:36:12 -0500 )edit

Don't exists this parameter in opencv_traincascade. It works?

The command line would look like this:

opencv_traincascade -data classifier -vec samples.vec -bg negatives.txt -numStages 20 -minHitRate 0.999 -maxFalseAlarmRate 0.5 -numPos 1300 -numNeg 1400 -w 48 -h 128 -mode ALL -precalcValBufSize 8192 -precalcIdxBufSize 8192 -featureType LBP setNumThreads(8)

Simão Martin gravatar imageSimão Martin ( 2017-06-02 14:08:16 -0500 )edit
LBerger gravatar imageLBerger ( 2017-06-02 14:23:41 -0500 )edit

Do not works. Continue using 64 threads spreading out to 64 cores. :(

Simão Martin gravatar imageSimão Martin ( 2017-06-05 09:24:30 -0500 )edit