I am currently on Windows and have spent quite a long time learning Eyesweb and it is a fab program.
My problem is that I would like to move completely to Linux, and they haven't made Eyesweb for Linux (annoying).
I've not tried video tracking in PD / GEM / PiDiP yet, and wondered if anyone could tell me how much can be done?
I am looking to do pattern tracking, i.e. track multiple objects that look the same. In Eyesweb this is called posture recognition. cv.jit for Jitter can do this as well.
I have just seen pdp_shape on the PiDiP website, which might be able to do the job. Can it track and identify blobs though? (computer tracking gurus will know what i mean)
Any advice in this area would be great thanks,
Hi Chris,
gem has a pix_blub object, you might use that in conjunction with pix_background to identify moving objects and their motion vectors. not sure how far this will get you, though ... I'm afraid pd /gem isn't as advanced as EyesWeb in this respect.
On the other hand, isn't EyesWeb open sourced, apart from the intel image processing libs? Maybe porting some stuff won't be too hard.
hope this helps, thoralf.
___________________________________________________________ Gesendet von Yahoo! Mail - Jetzt mit 250MB Speicher kostenlos - Hier anmelden: http://mail.yahoo.de
I was pretty sure EyesWeb is all closed source, but I'd love to be proven wrong.
I think PDP/PiDiP is the way to go for motion tracking. Currently it lacks EW's high level objects like the centroid tracker, but it has all the pieces there to build one.
The posture recognition in EW might use some sort of neural net to recognize postures, which might be buildable with Pd's ANN.
Let me know how it goes, as it's on my list to look at this.
-Ian
Thoralf Schulze wrote:
Hi Chris,
gem has a pix_blub object, you might use that in conjunction with pix_background to identify moving objects and their motion vectors. not sure how far this will get you, though ... I'm afraid pd /gem isn't as advanced as EyesWeb in this respect.
On the other hand, isn't EyesWeb open sourced, apart from the intel image processing libs? Maybe porting some stuff won't be too hard.
hope this helps, thoralf.
___________________________________________________________ Gesendet von Yahoo! Mail - Jetzt mit 250MB Speicher kostenlos - Hier anmelden: http://mail.yahoo.de
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://iem.at/cgi-bin/mailman/listinfo/pd-list
hi,
i've overestimated the sources-directories in the EywDevelopersTutorials.zip-package ... ian is right, eyesweb is closed source. the quote on their website: "Note: source code for various libraries is now provided as part of the distribution and is not available as a separate download." is a bit misleading: appearently only the sources of audio-related libraries are available to the public. bummer ...
thoralf.
___________________________________________________________ Gesendet von Yahoo! Mail - Jetzt mit 250MB Speicher kostenlos - Hier anmelden: http://mail.yahoo.de
Hi thanks for the replies,
Yeah unfortunately Eyesweb is closed source.
I've spent quite a long time with Eyesweb and it would be a shame to loose all that knowledge. I might have to think about using two machines, one with Windows and Eyesweb and one with Linux & PD, although this isn't an ideal solution.
Also I'm not sure if my 'Winfast VC 100 XP' capture card will run on Linux, but that's a separate issue really.
Also I can't try out PiDiP properly as I am on Windows, unless I use a Linux live cd, which is a pain.
I'd like to get away from Windows entirely, but not sure I can do what I need without Eyesweb.
Hmm, difficult to know what to do there.
Chris O'Shea
Ian Smith-Heisters wrote:
I was pretty sure EyesWeb is all closed source, but I'd love to be proven wrong.
I think PDP/PiDiP is the way to go for motion tracking. Currently it lacks EW's high level objects like the centroid tracker, but it has all the pieces there to build one.
The posture recognition in EW might use some sort of neural net to recognize postures, which might be buildable with Pd's ANN.
Let me know how it goes, as it's on my list to look at this.
-Ian
Le 16 Mars 2005 11:43, Chris O'Shea a écrit :
Also I'm not sure if my 'Winfast VC 100 XP' capture card will run on Linux, but that's a separate issue really.
It should work, as mentionned here: http://www.bttv-gallery.de/ -- Marc
I think your after largely the eyesweb feature of tracking multiple points of pressence/colour/motion?
I've talked to Mr Gridflow (Mathieu Bounchard) about this, as gridflow has pretty fast access to all pixels its exposed to. The reason why little software does it is that it is a complex problem! I personally can't get my head around the mathematics of how to make it work, but I know many have asked Mathieu and he is thinking about it.
So Mathieu, when will there be an example patch of tracking n number of points?
I'd love to be able to do this too!
pix_blob works very well for a single centroid, but thats it. pix_multiblob? ;)
B>
Chris O'Shea wrote:
Hi thanks for the replies,
Yeah unfortunately Eyesweb is closed source.
I've spent quite a long time with Eyesweb and it would be a shame to loose all that knowledge. I might have to think about using two machines, one with Windows and Eyesweb and one with Linux & PD, although this isn't an ideal solution.
Also I'm not sure if my 'Winfast VC 100 XP' capture card will run on Linux, but that's a separate issue really.
Also I can't try out PiDiP properly as I am on Windows, unless I use a Linux live cd, which is a pain.
I'd like to get away from Windows entirely, but not sure I can do what I need without Eyesweb.
Hmm, difficult to know what to do there.
Chris O'Shea
Ian Smith-Heisters wrote:
I was pretty sure EyesWeb is all closed source, but I'd love to be proven wrong.
I think PDP/PiDiP is the way to go for motion tracking. Currently it lacks EW's high level objects like the centroid tracker, but it has all the pieces there to build one.
The posture recognition in EW might use some sort of neural net to recognize postures, which might be buildable with Pd's ANN.
Let me know how it goes, as it's on my list to look at this.
-Ian
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://iem.at/cgi-bin/mailman/listinfo/pd-list
B. Bogart wrote:
pix_blob works very well for a single centroid, but thats it. pix_multiblob? ;)
oh darn. [pix_multiblob] (or [pix_blobtracker] as it was called i think) has been lying aroung here for approx. 1 year.
i don't know why i never put it into the cvs (but i always wanted to do so)
the only problem now is, that i cannot find the file. (next week a machine that was exposed at the cebit will hopefully come back and i am pretty sure that the code is there...)
thanks for remembering me
mfg.asd.r IOhannes
Hi Chris,
on a conference in Leeds the makers of Eyesweb told me that they were thinking of making a Linux version, maybe we should remind them :), it was last year. Has become PiDiP easier to install or is it still a struggle? I prefer GEM and tracking is there but basic, nothing in comparision to eyesweb.
Cheers,
Malte
Once you've installed PiDiP a dozen times or so it becomes straightforward. Let me see if I can summarize my procedure:
- Get source for PiDiP from Yves' site. The CVS is usually out of date.
- Get ffmpeg source FROM YVES' SITE. Note you can don't need to install
this source, you just need it to compile PiDiP. 3) Untar PiDiP and ffmpeg. 4) Configuration is tricky. You need to set --with-pd, --with-pdp and --with-ffmpeg to the directory where the base directories of their respective sources are. For me this is
PD_DIR = /home/ian/src/pure-data/pd PDP_DIR = /home/ian/src/pure-data/externals/pdp FFMPEG_SOURCE_DIR = /home/ian/src/pure-data/ffmpeg-cvs-2003-03-04
while pd is actually installed in /home/ian/pd.
- make && make install.
- cp pidip.pd_linux /path/to/pd/externs/ (make install doesn't do this)
- edit ~/.pdrc
Again, I think the only thing PiDiP lacks in comparison to EW is the high-level functions, but all the pieces are there to build them. With some clever bg-subtraction, masking, convolution, blob trackers, and averaging you could find the centroids. Then it's just a question of building some system for recognizing the postures, perhaps with ANN.
The EW developers might be responsive to at least giving tips/ pointing toward resources on how to do this. I guess it just depends how much time you have ;)
-Ian
Malte Steiner wrote:
Hi Chris,
on a conference in Leeds the makers of Eyesweb told me that they were thinking of making a Linux version, maybe we should remind them :), it was last year. Has become PiDiP easier to install or is it still a struggle? I prefer GEM and tracking is there but basic, nothing in comparision to eyesweb.
Cheers,
Malte
-
B. Bogart -
Chris O'Shea -
Ian Smith-Heisters -
IOhannes m zmoelnig -
Malte Steiner -
Marc Lavallée -
Thoralf Schulze