Ask Your Question

how opencv_traincascade can using GPU

asked 2017-12-05 23:58:24 -0600

Derick gravatar image

I have try to train my classifier using opencv_traincascade in my machine. When run it, it utilize 100% of my CPU but not use my GPU. I have install OpenCV 3.x in my Ubuntu 16.04. And I have GeForece GTX 1080 Ti/PCIe/SSE2. I success install the driver with CUDA 8.0. How I can use the GPU instead of using CPU? I use below script to train the module in terminal

opnencv_traincascade -data data -vec positives.vec -bg bg.txt -numPos 37000 -numNeg 756 -numStage 20 -w 20 -h 20

Any configuration I need to set to use the GPU?

edit retag flag offensive close merge delete



If you have a GTX1080Ti then skip the idea of cascade classifier all together and go for deep learning. Tons of efficient deep learning frameworks for object detection freely available.

StevenPuttemans gravatar imageStevenPuttemans ( 2017-12-06 04:27:07 -0600 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2017-12-06 02:11:37 -0600

berak gravatar image

sadly, there is no gpu code for the cascade training at all (it's not a configuration problem)

also, while there is cuda optimized detection code , you'd have to use the "old format" style cascades, generated from the opencv_haartraining tool (which is only in the outdated 2.4 branch)

edit flag offensive delete link more



And even then it seems that recent 2.4 branches have issues with the old models .... so better skip the idea alltogether :D

StevenPuttemans gravatar imageStevenPuttemans ( 2017-12-06 04:26:18 -0600 )edit

even I turn ON the CUDA during install OpenCV3.3, it still will not using GPU?

Derick gravatar imageDerick ( 2017-12-06 18:32:13 -0600 )edit

no way for the cascade training.

berak gravatar imageberak ( 2017-12-06 19:46:18 -0600 )edit

Question Tools

1 follower


Asked: 2017-12-05 23:58:24 -0600

Seen: 2,213 times

Last updated: Dec 06 '17