🚀 go-pugleaf

Deleos (deleos@gate.net) wrote:
: rmiller@bsm.bioc.ucl.ac.uk (Rob Miller) wrote:

: >Bill Leonard (bill_l@magicnet.net) wrote:
: >: In article <NEWTNews.813966557.30855.pts@pts.technet.sg>, pts@technet.sg wrote:

: >: > Hello There,
: >: > 
: >: >         We're looking for a system which can scan in the floor plans 
: >: > (preferably 2-D floor plans) using a scanner and build it into a 3-D output 
: >: > within half an hour.The 3-D output must also be able to cater for real-time 
: >: > walk through.

: >: It will also be at least 10 times more expensive than you're thinking, and
: >: require alot of custom programming work and years of development.

: >Naaaaw, 6 months for a working prototype.  Bet you could hack something to 
: >work on top of the DOOM engine even faster, tho I haven't worked with it...

: I would think that if _anyone_ could do it with the DooM engine id
: themselves could and they create their levels by hand.   Bill is
: _exactly_ right what is being proposed here is absurd for todays state
: of technology.  Obviously someone has seen to many science fiction
: movies. 

(1) Sorry to waste the bionet.software bandwidth, but it's the only group I 
monitor that has this thread; not too far off topic as protein folding is 
a 2-space to 3-space mapping problem as well.

(2) I was under the impression that the DOOM levels were based initially on, 
and defined as, 2D maps; as I said, I've not worked with it.

(3) I didn't say the time constraints would be met on a PC, I try to ignore 
the existence of PCs unless they're running unix.

(4) algorithm:

  (a) scan in the top view floor plans, do edge detection, find the corners.
      These are your initial x-y coords.  Specify a default height and you 
      can walk through right away, although everything in your model will 
      extend floor-to-ceiling.  Obviously the project is not feasible if 
      only top-view floor plans are given - but not because it `is absurd
      for today's state of technology,' rather because all the information is 
      not present.

  (b) scan in side view(s) to determine Z (height dimension in this case) coords 
      for everything in the rooms.  This part is trickier, but should be do-able 
      using constraint propagation and possibly restricting the 2D drawing format - 
      the original problem statement only says `preferably 2-D floor plans', so 
      defining some rules for linking various structures in the multiple drawings 
      is not out of the question.  

  There are many ways to specify 3D info in 2D, and at least some of them could 
be applied to this project -- software exists already to take a published stereo
pair drawing of a protein and convert that data into 3D coords, though not all 
that well I'm told.

(5) I haven't seen a decent science fiction film in years.


Internet: rmiller@bsm.bioc.ucl.ac.uk 
http://www.biochem.ucl.ac.uk/~rmiller

-----------------------------------------------------