[ros-users] Vision Stack Questions

[Abe Bachrach]

Hey Kurt,
this is probably on your guys' todo list for documentation/release, but
currently, there are a couple issues with the external files required by the
calonder features, and the place recognition...

calonder features:
the current.rtc file is not automatically downloaded when you build the
vslam_system package. To get it, I had to manually run
# make download_data_current.rtc
in the folder
stacks/visual_feature_detectors/calonder_descriptor/build

Furthermore, the path to the current.rtc file is hardcoded into the files
vision/vslam_system/run_stereo.cpp
vision/frame_common/src/frame.cpp


place recognition:
the path to the holidays.tree and holidays.weights is hardcoded  to
"/u/mihelich/vocab/holidays.tree", "/u/mihelich/vocab/holidays.weights"
in vslam.cpp (and possibly elsewhere)
these files don't seem to have a make rule that allows them to be
downloaded... are they available from somewhere?


finally, the software does not seem to check for the existence of these
files before trying to load them, and does not seem to check whether it was
able to successfully load them.

thanks,
-=Abe



On Wed, Jun 9, 2010 at 2:25 AM, Victor Eruhimov <relrotciv at googlemail.com>wrote:

> Hi, yes, this was my change causing errors -- I am in the process of
> integrating mono visual odometry. The compilation problem is fixed in
> r38197. Best Regards, Victor
>
> On Jun 9, 2010, at 9:40 AM, Kurt Konolige wrote:
>
> > Miguel, the VSLAM stack is experimental at this stage.  We'll try to have
> a fully documented release by the end of the summer.  At this point, there
> are four packages:
> >
> > sba - sparse bundle adjustment
> > posest - pose estimation, 3d and 2d
> > frame_common - front-end for camera parameters, matching
> > vslam_system - system-level integration
> >
> > These should compile correctly.  I just tried vslam_system and you're
> right, it doesn't - I've cc'd Victor who I think made the changes that
> caused the problem.
> >
> > Cheers --Kurt
> >
> > On 6/8/2010 5:18 AM, Miguel Juliá wrote:
> >>
> >> Hi all,
> >>
> >> I am trying to use your visual odometry code to test its implementation
> >> in the vision stack. What is the state of development of this stack? The
> >> documentation seems to be incomplete and there is no tutorial, just the
> >> code API. This stack includes visual odometry, place recognition and
> >> toro, doesn't it?
> >>
> >> I have downloaded the vision stack and tried to compile it but
> >> vslam_system return some compilation errors. It seems that the posest
> >> package in the repository is different from what vslam_system expects.
> >>
> >> Any advice?
> >>
> >> Thanks,
> >>
> >> Miguel Juliá.
> >>
> >>
> >>
> >>
> >> --
> >> Miguel Juliá Cristóbal
> >> Laboratorio de Automatización Robótica y Visión por Computador. (ARVC)
> >> Departamento de Ingeniería de Sistemas Industriales.
> >> Universidad Miguel Hernández.
> >> Av. Universidad s/n Edif. Quorum V.
> >> 03202 Elche-Alicante.
> >> Tlf.: 96522-2435
> >>
> >>
> >>
> >> _______________________________________________
> >> ros-users mailing list
> >> ros-users at code.ros.org
> >> https://code.ros.org/mailman/listinfo/ros-users
>
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20100609/8006e48b/attachment-0003.html>