An alternative to using Photoscan might be a multi-Kinect setup using
something like KScan
or Skanect or Microsoft´s SDK (as a start, I´m not sure multi-inputs are
supported as sample
code in the SDK).
That will seriously limit the achievable mesh detail due to the Kinext´s
sensor resolution
but the advantage is realtime capture since there is no need to
calculate depth.
Snapshots in realtime are possible. Not clean scans but fragmented pose
snapshots.
No shutter control thought...
I haven´t had the time to check out the Kinect One with any of the
above, which would
have a higher sensor resolution but also a different way of capturing
depth data that may
or may not work with your subjects.
I´ve read comments suggesting to add infrared light illumination to
improve quality off
depth capture but sofar haven´t tried that either.
For Photoscan, make sure to compare the Pro Version with the Standard
version.
You may need the Pro Version for the alignment features and Marker setup
it offers.
In any way, it can take minutes to hours to see a result of a capture,
which may add up
to too much time during a shoot.
Alternatives for photogrammetry based stuff are Nuke and PFTrack to
generate pointcloud data,
especially PFTrack can give great set data to work with if no LIDARs are
available.
I´ve seen collegues rebuild a location using multiple shots (scenes,
sequences) e.g. plates covering weeks
of shooting and then extending that set to taste, with all camera
positions tracked and aligned in worldspace
and real-world scale. Sounds tedious and overkill but actually saves
endeless amounts of guesswork, gives
great sense of scale and helps people work out lighting situations or
asset placements.
Related to capturing moving people the benefit is, you can have footage
made to align based on the environment,
even if it´s not in the captured set.
One problem I have with the standard version of Photoscan is that it´s
hard to force alignment to world and
generally real-world scales. That´s where the Pro version has more options.
Cheers,
tim
Am 29.01.2016 um 17:17 schrieb Chris Marshall:
I guessed that might be the case. I suppose I'm thinking realistically
double that for a full 360. Though we might only need to do 1 side.
Will try a few tests and make contact with fbfx to check out the costs
involved.
Thanks again
On 29 January 2016 at 16:10, <[email protected]
<mailto:[email protected]>> wrote:
you’d need more than 12 for a *good* 360 scan.. you could probably
just about achieve a 360 with 12 , but it would probably be quite
blobby and patchy.
We captured one side of a moving horse, with an array of 12,
genlocked 4k video cameras, and it wasn’t really enough, just for
doing one side. I know you’ll be shooting higher res than that,
but there are a lot of occlusions that 12 likely wont cover.
*From:* Chris Marshall <mailto:[email protected]>
*Sent:* Friday, January 29, 2016 3:41 PM
*To:* [email protected]
<mailto:[email protected]>
*Subject:* Re: photoscan on peaople
Thanks Adrian / Paul
It's that kind of stuff. A bit of action captured. I was thinking
we'd use about a dozen cameras, but I have a feeling it might need
to be many more. All synced, fast shutter, no motion blur etc.
We'd have a lot to do which is why I'm thinking of setting it up
ourselves. Just depends on the number of cameras that have the
right capabilities.
Cheers
On 29 January 2016 at 15:31, adrian wyer
<[email protected]
<mailto:[email protected]>> wrote:
we did some stuff a couple of years ago with fbfx at
shepperton studios, they were very friendly, and we got great
results... with people jumping around as if they'd been shot
(WWII stuff)
a
------------------------------------------------------------------------
*From:*[email protected]
<mailto:[email protected]>
[mailto:[email protected]
<mailto:[email protected]>] *On Behalf
Of *Rob Wuijster
*Sent:* 29 January 2016 15:27
*To:* [email protected]
<mailto:[email protected]>
*Subject:* Re: photoscan on peaople
If I'm correct, the guys at infinite realities
(http://ir-ltd.net/) use Agisoft for their full body scans.
Shooting for 3D models id a bit tedious at some points, but
there's a lot of examples to be found on Youtube and Vimeo.
To get really good 3D models, you need a LOT of photos from a
LOT of angles.
I'm not sure if moving people will work out for this stuff,
but am happily corrected if this is easily possible though.
Rob
\/-------------\/----------------\/
On 29-1-2016 16:13, Chris Marshall wrote:
Hi All,
Anyone used Agisoft Photoscan on people? We've not done much
of this kind of stuff, so any thoughts / feedback / pitfalls
etc are welcome. Or are there better solutions?
Thinking along the lines of capturing snapshots of moving
people with a bunch of synced cameras possibly.
Thanks
Chris
--
Chris Marshall
Mint Motion Limited
029 20 37 27 57
07730 533 115
www.mintmotion.co.uk <http://www.mintmotion.co.uk>
www.dot3d.com <http://www.dot3d.com>
Geen virus gevonden in dit bericht.
Gecontroleerd door AVG - www.avg.com <http://www.avg.com>
Versie: 2016.0.7357 / Virusdatabase: 4522/11509 - datum van
uitgifte: 01/29/16
--
Chris Marshall
Mint Motion Limited
029 20 37 27 57
07730 533 115
www.mintmotion.co.uk <http://www.mintmotion.co.uk>
www.dot3d.com <http://www.dot3d.com>
--
Chris Marshall
Mint Motion Limited
029 20 37 27 57
07730 533 115
www.mintmotion.co.uk <http://www.mintmotion.co.uk>
www.dot3d.com <http://www.dot3d.com>
