Forums > Jitter

cv.jit 1.6 beta 1 release

March 3, 2008 | 10:39 am

This is to let you know that I have released a new update to the cv.jit computer vision library.

http://www.iamas.ac.jp/~jovan02/cv/

Version 1.6 is now distributed as a true universal binary on OSX.

Apart from that, there are relatively few changes.

Two new externals have been added:

cv.jit.shift implements the MeanShift and CAMShift trackers which allow for relatively efficient and robust tracking of regions in a greyscale image.

cv.jit.resize, well, resizes. Since bicubic interpolation is used, down-sampled matrices will appear anti-aliased.

Some minor changes to existing objects:

There is now no limit to the number of points cv.jit.track can track at the same time. Of course, you should expect sluggish performance for very large number of points but 255 is no longer a hard-coded limit.

You can now query cv.jit.faces for the number of detected faces. This makes it easier to filter the dummy matrices that are output when no face has been detected.

cv.jit.features can now also compute features with sub-pixel accuracy for better tracking initialization.

On OSX, cv.jit.erode and cv.jit.dilate now use Core Image which means they now run much, much faster.

A number of bugs have also been fixed.

This first beta has been tested on Windows XP, Mac OS 10.4 with both Intel and PPC machines. It has not yet been tested on Vista or OS 10.5, though I do not think there should be any problem.

As always, I appreciate bug reports and comments!

Hope this will come useful.

Jean-Marc


March 3, 2008 | 2:40 pm

thank you Jean-Marc, always a pleasure to use cv.jit objects.
i’m also happy to see the cv.jit.faces which i couldn’t get to work on
windows seems to be working well now. cant wait to try it outside in the
real.

thanks again for the cv.jit packge which is one of the core attributes of
jittering.

yair

On Mon, Mar 3, 2008 at 12:39 PM, Jean-Marc Pelletier
wrote:

>
> This is to let you know that I have released a new update to the cv.jitcomputer vision library.
>
> http://www.iamas.ac.jp/~jovan02/cv/< http://www.iamas.ac.jp/%7Ejovan02/cv/>
>
> Version 1.6 is now distributed as a true universal binary on OSX.
>
> Apart from that, there are relatively few changes.
>
> Two new externals have been added:
>
> cv.jit.shift implements the MeanShift and CAMShift trackers which allow
> for relatively efficient and robust tracking of regions in a greyscale
> image.
>
> cv.jit.resize, well, resizes. Since bicubic interpolation is used,
> down-sampled matrices will appear anti-aliased.
>
> Some minor changes to existing objects:
>
> There is now no limit to the number of points cv.jit.track can track at
> the same time. Of course, you should expect sluggish performance for very
> large number of points but 255 is no longer a hard-coded limit.
>
> You can now query cv.jit.faces for the number of detected faces. This
> makes it easier to filter the dummy matrices that are output when no face
> has been detected.
>
> cv.jit.features can now also compute features with sub-pixel accuracy for
> better tracking initialization.
>
> On OSX, cv.jit.erode and cv.jit.dilate now use Core Image which means they
> now run much, much faster.
>
> A number of bugs have also been fixed.
>
> This first beta has been tested on Windows XP, Mac OS 10.4 with both Intel
> and PPC machines. It has not yet been tested on Vista or OS 10.5, though I
> do not think there should be any problem.
>
> As always, I appreciate bug reports and comments!
>
> Hope this will come useful.
>
>
> Jean-Marc
>


March 3, 2008 | 4:42 pm

a quick one, this patch will down-res a 640×480 camera grab to 80×60, detect
a face and output a face using the higher res.

one minor problem, on windows i had to add the cv.jit-supportdata subdir to
the search dir as cv.jit.faces faild to find the detection model by itself.

this is for windows, to make it work on mac change the [jit.dx.grab] to [
jit.qt.grab]

#P window setfont "Sans Serif" 9.;
#P window linecount 4;
#P comment 197 407 152 9109513 caution windows users- add cv.jit-support/DATA
subdir to your file prefernnces as cv.jit.faces doesnt seem to know about
it.;
#P window linecount 1;
#P newex 238 364 26 9109513 print;
#P message 44 340 159 9109513 srcdimstart $1 $2 , srcdimend $3 $4;
#P newex 219 314 56 9109513 vexpr $i1*8;
#P newex 219 270 80 9109513 jit.matrix @thru 0;
#P message 103 55 30 9109513 close;
#P newex 219 291 40 9109513 jit.iter;
#P newex 149 243 28 9109513 sel 1;
#P newex 131 221 46 9109513 change 1;
#P message 123 146 53 9109513 getnfaces;
#P newex 131 199 66 9109513 route nfaces;
#P newex 27 386 43 9109513 qlim 250;
#P user jit.pwindow 26 407 162 122 0 0 0 0 1 0;
#P message 73 55 28 9109513 open;
#P newex 219 124 66 9109513 jit.rgb2luma;
#P newex 27 364 208 9109513 jit.matrix 4 char 640 480 @thru 0 @usesrcdim 1;
#P newex 27 97 203 9109513 t l b l;
#P newex 319 228 43 9109513 qlim 250;
#P newex 319 207 89 9109513 cv.jit.faces.draw;
#P newex 219 146 142 9109513 jit.matrix 1 char 80 60 @adapt 0;
#P user jit.pwindow 318 248 82 62 0 0 0 0 1 0;
#P newex 219 171 62 9109513 cv.jit.faces;
#P toggle 26 37 15 0;
#P newex 26 55 46 9109513 metro 2;
#P newex 26 76 180 9109513 jit.dx.grab 640 480 @unique 1 @format 4;
#P window setfont "Sans Serif" 12.;
#P comment 144 44 163 9109516 output matrix if face detected;
#P window setfont "Sans Serif" 9.;
#P comment 211 78 142 9109513 fire-i Y411 , 640×480;
#P fasten 24 0 11 0 49 361 32 361;
#P fasten 24 0 25 0 49 361 243 361;
#P connect 7 0 5 0;
#P fasten 7 0 8 1 224 167 403 167;
#P connect 9 0 6 0;
#P connect 8 0 9 0;
#P connect 5 0 22 0;
#P fasten 5 0 8 0 224 201 324 201;
#P fasten 20 0 23 0 224 310;
#P fasten 22 0 20 0 224 291 224 291;
#P fasten 19 0 11 0 154 267 32 267;
#P fasten 19 0 22 0 154 267 224 267;
#P fasten 17 0 5 0 128 168 224 168;
#P connect 12 0 7 0;
#P connect 10 2 12 0;
#P connect 18 1 19 0;
#P fasten 16 0 18 0 136 221 136 221;
#P fasten 5 1 16 0 276 195 136 195;
#P connect 10 1 17 0;
#P fasten 23 0 24 0 224 336 49 336;
#P connect 15 0 14 0;
#P connect 11 0 15 0;
#P connect 10 0 11 0;
#P connect 2 0 10 0;
#P fasten 21 0 2 0 108 75 31 75;
#P fasten 13 0 2 0 78 75 31 75;
#P connect 3 0 2 0;
#P connect 4 0 3 0;
#P window clipboard copycount 27;

On Mon, Mar 3, 2008 at 4:40 PM, yair reshef wrote:

> thank you Jean-Marc, always a pleasure to use cv.jit objects.
> i’m also happy to see the cv.jit.faces which i couldn’t get to work on
> windows seems to be working well now. cant wait to try it outside in the
> real.
>
> thanks again for the cv.jit packge which is one of the core attributes of
> jittering.
>
> yair
>
>
> On Mon, Mar 3, 2008 at 12:39 PM, Jean-Marc Pelletier
> wrote:
>
> >
> > This is to let you know that I have released a new update to the cv.jitcomputer vision library.
> >
> > http://www.iamas.ac.jp/~jovan02/cv/< http://www.iamas.ac.jp/%7Ejovan02/cv/>
> >
> > Version 1.6 is now distributed as a true universal binary on OSX.
> >
> > Apart from that, there are relatively few changes.
> >
> > Two new externals have been added:
> >
> > cv.jit.shift implements the MeanShift and CAMShift trackers which allow
> > for relatively efficient and robust tracking of regions in a greyscale
> > image.
> >
> > cv.jit.resize, well, resizes. Since bicubic interpolation is used,
> > down-sampled matrices will appear anti-aliased.
> >
> > Some minor changes to existing objects:
> >
> > There is now no limit to the number of points cv.jit.track can track at
> > the same time. Of course, you should expect sluggish performance for very
> > large number of points but 255 is no longer a hard-coded limit.
> >
> > You can now query cv.jit.faces for the number of detected faces. This
> > makes it easier to filter the dummy matrices that are output when no face
> > has been detected.
> >
> > cv.jit.features can now also compute features with sub-pixel accuracy
> > for better tracking initialization.
> >
> > On OSX, cv.jit.erode and cv.jit.dilate now use Core Image which means
> > they now run much, much faster.
> >
> > A number of bugs have also been fixed.
> >
> > This first beta has been tested on Windows XP, Mac OS 10.4 with both
> > Intel and PPC machines. It has not yet been tested on Vista or OS 10.5,
> > though I do not think there should be any problem.
> >
> > As always, I appreciate bug reports and comments!
> >
> > Hope this will come useful.
> >
> >
> > Jean-Marc
> >
>
>


March 4, 2008 | 3:51 am

> one minor problem, on windows i had to add the cv.jit-supportdata subdir to
> the search dir as cv.jit.faces faild to find the detection model by itself.

Really? I tried it on two different computers and I can’t seem to be able to reproduce the problem. The models load fine. In the cv.jit.blobs.recon help file can you load the training data properly? Also, can you read the test images for the cv.jit.blobs.*** objects (arrows.gif and abstract.jpg)?


March 4, 2008 | 7:41 am

ok, my bad. it works without specifying the path. probably didn’t restart
max or something.
was too excited ;)

On Tue, Mar 4, 2008 at 5:51 AM, Jean-Marc Pelletier
wrote:

>
> > one minor problem, on windows i had to add the cv.jit-supportdatasubdir to
> > the search dir as cv.jit.faces faild to find the detection model by
> itself.
>
> Really? I tried it on two different computers and I can’t seem to be able
> to reproduce the problem. The models load fine. In the cv.jit.blobs.reconhelp file can you load the training data properly? Also, can you read the
> test images for the cv.jit.blobs.*** objects (arrows.gif and abstract.jpg
> )?
>


March 15, 2008 | 3:35 pm

Hi,

I don’t seem to get the cv.jit package fully to work.
I use xp and Jitter 1.6.3
some work some don’t.
these are the errormessages it produces:
• error: Error 126 loading external cv.jit.erode
• error: Error 126 loading external cv.jit.dilate

any help appreciated

Tim


March 17, 2008 | 9:27 am

I’m terribly sorry about that. The wrong zip file found its way on the server when I did minor maintenance. Download everything again, it should work now.

Apologies for the inconvenience.

Jean-Marc

Quote: Tim van Elferen wrote on Sat, 15 March 2008 09:35
—————————————————-
> Hi,
>
> I don’t seem to get the cv.jit package fully to work.
> I use xp and Jitter 1.6.3


Viewing 7 posts - 1 through 7 (of 7 total)