[ExI] pokemon go further
outlawpoet at gmail.com
Fri Jul 15 16:10:05 UTC 2016
Lockheed Martin is experimenting with technology from NGRAIN for this
purpose in the F-35 assembly line(which may not be the best example, given
the cost overruns associated with that program).
NGRAIN was using Epson Moverios(head mounted displays with sensors) early
last year, but they may have moved to a new platform like HoloLens. They
have an Android public demo you can find here:
As Adrian alluded to above, this is more about the scut work to
characterize and library all the necessary procedures and objects in
sufficient detail, the sensors and cv is probably up to the task (although
as the number of objects possible to recognize between goes up, errors go
up as well, unless intelligent error correction and pruning is used).
On Tue, Jul 12, 2016 at 5:11 PM, Adrian Tymes <atymes at gmail.com> wrote:
> On Jul 12, 2016 4:49 PM, "spike" <spike66 at att.net> wrote:
> > This game gets partway there to something I have been anticipating for a
> long time: some kind of virtual reality in which characters or something is
> superimposed on the scene to create a VR.
> It's called "augmented reality". People have been playing with it for
> decades, though this seems to be the most widespread single app deployment
> so far, by a wide margin.
> > An example I have used before to describe this is in repairing the
> secondary gear case of a Suzuki Cavalcade. Since we know exactly what that
> gear case looks like, the image recognition task is limited to a workable
> scale. When a phone is held about a meter from that gearcase, the software
> would recognize it and point to the four bolts which must be removed to get
> to the next step. It could tell you the size wrench needed, the torque
> value on reassembly, etc.
> > Have we any image-recognition gurus or semi-hipsters among us who can
> comment on how difficult is this problem? I am thinking of a very limited,
> closed-ended image recognition subroutine which would recognize the outline
> of a known quantity, such as a specific motorcycle part. Software can
> recognize faces; this seems trivially easy in comparison. It seems like
> just a scale and rotate problem to get what the camera sees to match what
> it is looking for.
> This is the Maintenance Aid Computer concept that Boeing attempted to
> deploy (with special goggles and gloves) for aircraft in the '90s. It
> would probably be possible to walk people through certain specific
> repairs. The challenge might be more in building up a sufficient library
> of repairs (which is more laborious than inventive, since you have to code
> up one repair, then the next, and so on...unless you came up with good
> tools to add to the library too) to be more useful than a gimmick.
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
outlawpoet at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the extropy-chat