[Bf-committers] GSoC 2016: Finish multi-camera reconstruction

[Sergey Sharybin]

If you aren't familiar with such VFX workflow, researching this a bit will
be a good starting point. Can't give exact links for that, but it's easy to
google and find example footage.

Next step would be to become more familiar with what's happening in
Blender. Briefly:

- intern/libmv -- is the bundled version of Libmv in Blender, including
C-API
- source/blender/blenkernel/intern/tracking* -- is the lower-level
implementation of all tracking related stuff in Blender
- source/blender/editor/space_clip* -- is the interface for the tracking in
Blender (with some operators to handle the data etc).

Then it worth checking on the autotracking API in Libmv. I can't remember
if we moved it from Blender to Libmv upstream, but that's something we'll
be switching to in the future. It has some basement for supporting multiple
clips, but it wasn't finished, needs some review and hook up to Blender.

This project will also involve some nice an friendly tool implemented to
make it easy to operate with multiple clips from the interface. This is
something where artists input will be really welcome. We've got Sean
Kennedy and Sebastian Koenig aboard for this. Sebastian is usually in IRC
and easily reachable ;)

Hope this covers overall directions, i'll be happy to answer more questions
in details if you get any :)


On Wed, Mar 2, 2016 at 12:21 PM, Tianwei Shen <shentianweipku at gmail.com>
wrote:

> Hi Sergey,
>
> Thanks for your guidance and clarifications. I’ve successfully compiled
> the blender code. I am now getting acquittance with the project structure.
> Do you have any suggestion, like where I should start, if I want to work on
> the multi-camera reconstruction stuff? Thanks.
>
>
> Best,
> Tianwei
>
> > On Mar 2, 2016, at 3:16 PM, Sergey Sharybin <sergey.vfx at gmail.com>
> wrote:
> >
> > Hi,
> >
> > It is definitely Structure-from-Motion, was just a typo. Fixed now.
> >
> > Term "witness camera" is a commonly used in VFX industry. Briefly: the
> > scene is being shot by multiple cameras, one of the cameras are used for
> > the finals (footage from it goes to a final film), others are only used
> to
> > gather more information for solver to avoid ambiguities.
> >
> > For getting Blender compiled, did you check
> > http://wiki.blender.org/index.php/Dev:Doc/Building_Blender ?
> >
> >
> > On Tue, Mar 1, 2016 at 9:50 PM, Tianwei Shen <shentianweipku at gmail.com>
> > wrote:
> >
> >> Hi,
> >>
> >> I am Tianwei Shen, a second-year Ph.D. student in HKUST, working mainly
> on
> >> image-based modeling. I am interested in the GSoC topic 'Finish
> >> multi-camera reconstruction' and have some question related to it.
> >> Specifically I don’t quite understand terms like ‘witness cameras’ and
> >> ‘source-from-motion’. Maybe we use different terminology so in my view
> >> ‘source-from-motion’ may be similar to Structure-from-Motion, isn’t it?
> I
> >> notice that blender relies on libmv on the motion tracking part. I also
> >> have experience working with this library for registering the poses of
> >> cameras. I am now trying to compile this project and would like to have
> a
> >> discussion with you in greater depth. Can anyone guide me where to
> start?
> >> Thanks a lot.
> >>
> >>
> >> Thanks,
> >> Tianwei
> >> _______________________________________________
> >> Bf-committers mailing list
> >> Bf-committers at blender.org
> >> http://lists.blender.org/mailman/listinfo/bf-committers
> >>
> >
> >
> >
> > --
> > With best regards, Sergey Sharybin
> > _______________________________________________
> > Bf-committers mailing list
> > Bf-committers at blender.org
> > http://lists.blender.org/mailman/listinfo/bf-committers
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>



-- 
With best regards, Sergey Sharybin