Subject:
|
Re: Whose angel sculpture was that?
|
Newsgroups:
|
lugnet.events.brickfest
|
Date:
|
Mon, 29 Aug 2005 20:49:32 GMT
|
Viewed:
|
1751 times
|
| |
|
|
In lugnet.events.brickfest, Jordan Bradford wrote:
| |
In lugnet.events.brickfest, David Winkler wrote:
| |
In lugnet.events.brickfest, Jordan Bradford wrote:
| |
In lugnet.events.brickfest, Wayne R. Hussey wrote:
| |
In lugnet.events.brickfest, Joe Strout wrote:
| |
There was a life-sized angel statue in the Sculpture room. I read the
card, but failed to write down (or remember) the name of the builder.
Does anyone know who that was, and how to contact him?
I’d like to put a picture of it in BrickWiki (on the Sculpture page), but
I need permission first.
Thanks,
- Joe
|
That Angel was created by David Winkler, who is also our NWBrickCon 2005
Art/Sculpture/Mosaic Coordinator. If you wish to contact him, fill out and
send the “Contact Us” link at
NWBrickCon and we’ll help you connect.
Wayne
|
Wasn’t that sculpture built as a test of some software he had written? I
believe you give it a 3D model and it generates building instructions for
you.
|
Yup.
I gave a talk on the software at Brickfest, “Automated Brick Layout”
I posted the slides from that talk in my brickshelf folder:
http://www.brickshelf.com/gallery/happyfrosh/BrickFest2005/automatedbricklayout.pdf
Please feel free to post photos.
Thank you!
-dw
|
I missed your presentation that day. How technical was it? You probably
explained NP-complete briefly, and I’m guessing you didn’t go into the
computer science part of everything except what was necessary. I’m curious if
you have a more technical version of a presentation I could see -- I’d at
least like to know how the lossy/lossless algorithms work in psuedo code.
Also, is this subject worthy of an ACM article? That would be cool.
Thanks!
|
NP-complete is a class of difficult
problems in Computer Science. In general it’s a problem too difficult to be
solved exactly. A well-known example is the
traveling salesman
problem. In my talk I sort of glossed over this.
While I had originally intended the work to be a SigGraph paper, time constraints prevented. Siggraph likes things to have
never been presented elsewhere, and BrickFest would likely count. So my chances
of a paper there are fairly slim.
From the
talk: • Lossy techniques that work – Discard some high cost states
• This is called a Beam
Search
Basicly what this boils down to...
To start, the region of size 0 is just the region with no legos added.
To fill a region of size m:
Find a region of size m-n, and add a piece of size n. Check whether that exact
set of covered squares has already been found. If so, take the one with the
lower cost. Here’s the lossy part: If it’s not in the top w best scores for
that size, throw it out. The “exact set of covered squares” is just a bitfield
of the size of the smallest covering rectangle. These are stored in a binary
tree for fast searching (I used the STL implementation).
For each number of squares filled in the region this is repeated for all n.
For example: to find the regions of size 3, we have to look at the regions of
size 0 (the empty region) with a lego of area 3 added, regions of size 1 (one
square covered) adding a lego of area 2 added (in each of the legos
orientations), and a region of size 2 with a lego of area 1 (a 1x1) added.
Where w is the width of the beam. In general I use w=2000 for quick renders,
and w=20000 for final renders.
At the end choose the path that has the lowest cost.
This is basicly searching the lattice shown in the presentation from the top of
the page to the bottom of the page.
I’ll work on better pseudo-code.
-dw
|
|
Message is in Reply To:
 | | Re: Whose angel sculpture was that?
|
| (...) I missed your presentation that day. How technical was it? You probably explained NP-complete briefly, and I'm guessing you didn't go into the computer science part of everything except what was necessary. I'm curious if you have a more (...) (19 years ago, 26-Aug-05, to lugnet.events.brickfest, FTX)
|
11 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
