Jean-Marc Pelletier

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

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

Apart from that, there are relatively few changes.

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.

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.

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!

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-1-6-beta-1-release

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

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

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

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

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

> Apart from that, there are relatively few changes.

> cv.jit.shift implements the MeanShift and CAMShift trackers which allow

> for relatively efficient and robust tracking of regions in a greyscale

> 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

> cv.jit.features can now also compute features with sub-pixel accuracy for

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

> 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!

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.

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/
>
> 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
>

a quick one, this patch will down-res a 640x480 camera grab to 80x60, 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 [

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

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

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

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

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

> > Apart from that, there are relatively few changes.

> > cv.jit.shift implements the MeanShift and CAMShift trackers which allow

> > for relatively efficient and robust tracking of regions in a greyscale

> > 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

> > cv.jit.features can now also compute features with sub-pixel accuracy

> > On OSX, cv.jit.erode and cv.jit.dilate now use Core Image which means

> > A number of bugs have also been fixed.

a quick one, this patch will down-res a 640x480 camera grab to 80x60, 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]

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/
> >
> > 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
> >
>
>

> 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)?

> 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)?


ok, my bad. it works without specifying the path. probably didn't restart

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

> > 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

> 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

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
> )?
>

I don't seem to get the cv.jit package fully to work.

• error: Error 126 loading external cv.jit.erode

• error: Error 126 loading external cv.jit.dilate

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

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.

Quote: Tim van Elferen wrote on Sat, 15 March 2008 09:35

----------------------------------------------------

> I don't seem to get the cv.jit package fully to work.

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


cv.jit 1.6 beta 1 release