Ask Your Question

when does haar function not work?

asked 2015-04-13 11:34:16 -0500

cellurl gravatar image


When does the Haar function not work?

EG, if I do a digital zoom, will it work? EG, if I have a big face, will it work?

Sorry to ask this, but I am lazy (there I said it).....


edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2015-04-14 04:19:05 -0500

Your laziness doesn't help. Please improve your question more clearly. Are you talking about the performance of the cascade of weak classifier implementation?

  • Digital zoom --> depends on the amount of artefacts you create ... but keeping in mind that a normal model is 24x24 pixels large, you will capture most of them
  • Big face = no problem due to the scale pyramid. Your model defines the smallest possible scale detectable. Larger faces are no problem at all.
edit flag offensive delete link more


soo, if I want to capture more distant cars with the same classifier, I would grab pixels just in the middle and since that is less than the pyramid-smallest, I would have to add (interleave, filter, upsample) to get more pixels between them? Does that make sense? Thanks Steven!



cellurl gravatar imagecellurl ( 2015-04-16 08:51:54 -0500 )edit

No, you would build your pyramid so that your smallest object size, the object model dimensions, is the smallest object that you want to track! The algorithm and the detectMultiScale takes care of all the rest!

StevenPuttemans gravatar imageStevenPuttemans ( 2015-04-17 03:23:57 -0500 )edit

Question Tools

1 follower


Asked: 2015-04-13 11:34:16 -0500

Seen: 91 times

Last updated: Apr 14 '15