Re: unigrafix "escapes" for space labels, room numbers

[Jason Bell <bellj@graphics.lcs.mit.edu>]

Allen Miu's code generates a simple file with space labels from the
Autocad files.  The executable is here:

/walkthru/SLAM/getrooms

and the output is a simple x,y,z, labelname, spaceType

-Jason

Seth Teller wrote:

> sean, will,
>
> here's a suggestion about how we can generate space
> labels, room numbers, etc. in the high-LOD floorplans.
> what i have in mind here is 3D text on or near
> relevant doors, and/or floating at the center of
> named spaces.
>
> it would be relatively easy to process the autocad
> layer containing space labels, and convert each to
> a unigrafix "escape" sequence, see
>
>   http://graphics.lcs.mit.edu/~seth/datasets/ug.fileformat
>
> this is just a left paren, followed by some literal text,
> and a right paren.  i propose that you extend ug2iv to
> recognize such statements, for example inside ug defs,
> and produce inventor text nodes as output, e.g.,
>
>   Text3 { string "label text" justification CENTER }
>
> assuming that the text location is propagated correctly
> by acad2ug, this should result in text labels in the
> generated ug (and thus iv and wrl files).  note that this
> places space labels near the spaces they describe, but
> there is no *semantic* binding retained between spaces
> and labels.
>
> so:  even better, in the longer run, would be to associate
> name labels with named spaces, explicitly.  this could be
> done, for example, by wrapping the space's defining
> geometry, along with its label(s) and other metadata, in
> an enclosing (perhaps analogously named) ug "def".
> describe, not just geometrically.
>
> it turns out that allen miu (in code now taken over by jason
> bell, cc'ed) worked out how to associate space labels with
> spaces in the source floorplans.  for each space-label string
> his code encounters, it finds the centroid of the string, and
> finds the space boundary contour that encloses the centroid.
> (to handle big labels placed near small spaces, i think his
> code searches for the indicator line segment and binds the
> label to  the space enclosing the other end of the segment.)
>
> thus, we have the infrastructure in place to do an intelligent
> job of binding geometry and names.  let's talk about how we
> can get the benefit of this in our continued elaboration of
> procedurally-generated MIT models.
>
> prof. t.