slow tracking

[shay]

Hi, i'm using ctrax on a video with only one fly, since it didn't
accept the coded videos I loaded an uncompressed video (~2G per
minute), the tracking takes about X12 of real time (12 minutes for
tracking one minute of video) is that normal? if not, how can i speed
it up? I also tried to track an sbfmf file and it took about the same
time but when I loaded the example videos from the site it was much
faster (about 100 seconds for tracking 1 minute of video).

thank you, shay.

[shay]

I've seen that Georg posted about a similar problem but since he was
tracking 40 flies I guess it's not completely the same.

[Kristin Branson]

If the other video is tracking faster, then it sounds like you're not setting the parameters in the ideal way (this is currently something of an art, and unfortunately requires a little bit of understanding of the tracking algorithm). The algorithm is slow when trying to split or merge clumps of pixels into one or multiple flies. See http://www.dickinson.caltech.edu/ctrax#ctrax-usage for documentation.

Kristin

[shay]

First of all I'd really like to thank you for all the help.
now, I've misled you before, the example videos do not track faster, I
think that the first time I tried them I used the existing annotation
file and that's the reason I thought it was faster, they actually go
much slower (1 frame per second for the 10 flies video).
I still think that the fact that I'm only tracking one fly should give
me (with the right parameters being set) a great advantage in terms of
tracking speed but even after reading the Ctrax usage part and a bit
of trial and error I only got to about 10-11 minutes for tracking one
minute of video (and even that was after eliminating all the hindsight
options which is not a good idea...).
Do you have any tips on parameters I should drastically change because
I'm only tracking one fly?
Also, what was the best tracking speed you had with Ctrax and with how
many flies tracked?

Thanks again,

Shay

On Sep 23, 5:39 pm, Kristin Branson <kristinbran...@gmail.com> wrote:
> If the other video is tracking faster, then it sounds like you're not
> setting the parameters in the ideal way (this is currently something of an
> art, and unfortunately requires a little bit of understanding of the
> tracking algorithm). The algorithm is slow when trying to split or merge

> clumps of pixels into one or multiple flies. Seehttp://www.dickinson.caltech.edu/ctrax#ctrax-usagefor documentation.

>
> Kristin
>
>
>
> On Tue, Sep 22, 2009 at 7:27 AM, shay <snoopyco...@gmail.com> wrote:
>
> > Hi, i'm using ctrax on a video with only one fly, since it didn't
> > accept the coded videos I loaded an uncompressed video (~2G per
> > minute), the tracking takes about X12 of real time (12 minutes for
> > tracking one minute of video) is that normal? if not, how can i speed
> > it up? I also tried to track an sbfmf file and it took about the same
> > time but when I loaded the example videos from the site it was much
> > faster (about 100 seconds for tracking 1 minute of video).
>

> > thank you, shay.- Hide quoted text -
>
> - Show quoted text -

[shay]

Oh, one more thing, I know this is a long shot but do you think it has
something to do with the fact that in my video the fly is dark and the
background is bright? if i'll make a negative of that video do you
think it will make any difference?

> > - Show quoted text -- Hide quoted text -