Softcascades for Human Detection

asked 2014-07-01 02:43:49 -0600

yohanrw gravatar image


I need to do human motion detection on real time. I heard OpenCV 3.0 comes with Softcascades trained for humans as well. I also heared it is very fast. However I did not find any article or sample in any web page regarding this. From where can I get the softcascades for real time human detection? Or can you give me some info?

Apart from that, I must run this detector using multiple cameras. That means, I have 4-5 cams (for now) and I need to run the detector in real time in video feeds generated by all the cams. So I can watch for people using all the cams. In order to do this, how fast the computer should be? any idea?

I really appreciate your kind advices here because I have to make an investment on computers etc.

edit retag flag offensive close merge delete


@berak: is it just moved or really gone, somehow my browser doesn't want to load this github page...

Guanta gravatar imageGuanta ( 2014-07-01 05:44:56 -0600 )edit

@Guanta, awfully sorry. did not realize it was the full commit diffs ;(

berak gravatar imageberak ( 2014-07-01 06:20:01 -0600 )edit

(sorry, for making spagetti of this thread, but i had to delete prev.comment)

so, the softcascade was removed a few days ago.

berak gravatar imageberak ( 2014-07-01 06:26:50 -0600 )edit

hmm okay, let's see if it comes back in contrib sometime.

Guanta gravatar imageGuanta ( 2014-07-01 06:48:04 -0600 )edit

@berak: OK, so have any suggestions about my question?

yohanrw gravatar imageyohanrw ( 2014-07-01 11:30:21 -0600 )edit

@berak: You mean the softcascades are not supported any more?

yohanrw gravatar imageyohanrw ( 2014-07-01 11:33:20 -0600 )edit

that probably means more : in transit, in the process of getting moved to the contrib repo (at least i read it this way.) .

but git never forgets, you can just checkout the commit just before that.

if you already got 3.0, you might as well try it

oh, it seems to be patented

berak gravatar imageberak ( 2014-07-01 11:35:48 -0600 )edit