[Bf-committers] Blender At Your Fingertips: Prototype Unistroke Commands for Tablets

[Jason Wilkins]

More details about the video and the prototype.

The recognizer used in the video is very simple to implement and
understand.  It is called $1 (One Dollar) and was developed at the
University of Washington [1].  We had a seminar recently about
interfaces for children and extensions to $1 were presented and I was
inspired by their simplicity because it meant I could just jump right
in.  It works OK and is good enough for research purposes.

One thing $1 does not do is input segmentation.  That means it cannot
tell you how to split the input stream into chunks for individual
recognition.  What I'm doing right now is segmenting by velocity.  If
the cursor stops for 1/4 of a second then I attempt to match the
input.  This worked great for mice but not at all for pens due to
noise, so instead of requiring the cursor to stop I just require it to
slow down a lot.  I'm experimenting with lots of different ideas in
rejecting bad input.  I'm leaning towards a multi-modal approach where
every symbol has its own separate criteria instead of attempting a
one-size-fits-all approach.

The recognizer is driven by the window manager and does not require a
large amount of changes to capture the information it needs.
Different recognizers could be plugged into the interface.

The "afterglow" overlay is intended to give important feedback about
how well the user is entering commands and to help them learn.  The
afterglow gives an indication that a command was successfully entered
(although I haven't disabled the display of valid but unbound gestures
yet).  The afterglow morphs into the template shape to give the user
both a clearer idea of what the gesture was and to help the user fix
any problems with their form.

In the future I want to use information about the gesture itself, such
as its size and centroid, to drive any operator that is called.  For
example, drawing a circle on an object might stamp it with a texture
whose position and size were determined by the size and position of
the circle.

Additionally I want to create a new window region type for managing,
training, and using gestures.  That might be doable as an add-on.

[1] https://depts.washington.edu/aimgroup/proj/dollar/


On Sun, Nov 18, 2012 at 7:42 AM, Jason Wilkins
<jason.a.wilkins at gmail.com> wrote:
> I've been exploring some research ideas (for university) and using
> Blender to prototype them.  I made a short video that demonstrates
> what I was able to do the last couple of days.  I'm starting to create
> a general framework for sketch recognition in Blender.
>
> http://youtu.be/IeNjNbTz4CI
>
> The goal is an interface that could work without a keyboard or most
> buttons.  I think a Blender with gestures is far more like Blender
> than a Blender that is plastered with big buttons so it works on a
> tablet.  It puts everything at your fingertips.