Can HAAR classifier be used between two different camera modules

asked 2014-04-20 03:37:34 -0500

DoX gravatar image

If I train a HAAR classifier using one camera, say on an iPhone using OpenCV. Can I export that HAAR classifier to another device with a different camera module. My thought is a user could generate HAAR metadata on an iPhone, then load that into an OpenCV implementation running on an embedded Linux system (like cubieboard, or a Ubunu system w/ a webcam or other camera module).

Is that HAAR metadata camera module dependent?

edit retag flag offensive close merge delete


you don't want to train any kind of cascade on your phone. ;) it usually needs some hours on a beefy machine with lots of ram.

also, the training images should not show any dependancy on the device used to take them, and the resulting cascade-data is not dependant on the current machine.

(is that , what you meant with 'metadata' even ?)

berak gravatar imageberak ( 2014-04-20 03:57:20 -0500 )edit

Thanks for your answer. Is your recommendation against training on a mobile device due to its computational constraints or due to the camera module specs in most phones? I would like to capture training images on a small RBG camera module, then send these images to a larger device (with beefy CPU/GPU & memory) to generate the cascade data. I can control the small camera module with high precision and repeatability (it is on an arm of a robot).

DoX gravatar imageDoX ( 2014-04-21 20:58:40 -0500 )edit