Impressed by the recent knot images posted recently on this newsgroup by 
Bill Pragnell and others, I decided to try something in this direction 
too. The knot shown here is Ashley knot #2334 (see 
http://www.cs.ubc.ca/nest/imager/contributions/scharein/Ashley/).

The circuit was done entirely with Wings3D, using as a guide a model 
exported from knotplot. The grass texture is taken from Rune's website, 
and the car is nothing but G. Tran's mini.

All comments are welcome. I am particurlary interested by ideas about 
the background: I have tried a few things without much success.

Thibaut

Post a reply to this message

Attachments:
Download 'circuitd_33.jpg' (163 KB)

Preview of image 'circuitd_33.jpg'

I'm thinking of a very tedious and tiresome (no puin intended) animation as 
viewed from the drivers seat of one of those cars as it traverses the entire 
knot.  Only joking. Very ingenious use of a knot. While I wondered if two 
lanes might have been better for traffic I followed it around and found you 
have all the cars going in the same direction, so guess that'll be okay.  ;)

Now I must ask... Were you able to place the cars using the same method the 
road is done? I figure they are, somehow, but being unevenly spaced that 
might not be so, so I am asking please. Got to know.  :)

-- 
Bob H  www.3digitaleyes.com
http://3digitaleyes.com/imagery/

Post a reply to this message

> knot.  Only joking. Very ingenious use of a knot. While I wondered if two 
> lanes might have been better for traffic I followed it around and found you 
> have all the cars going in the same direction, so guess that'll be okay.  ;)

My idea at first was with two opposite traffic lines, but this makes car 
  two times smaller on the picture, which is somewhat problematic.

> 
> Now I must ask... Were you able to place the cars using the same method the 
> road is done? I figure they are, somehow, but being unevenly spaced that 
> might not be so, so I am asking please. Got to know.  :)
> 

The cars are simply placed in pov, using position and angle infos read 
in Wings3D. I just chose the points where I wanted the cars to be, read 
the corresponding position (and angles) in Wings, and then 
translate/rotate the car object as needed in pov. I wish a had the 
spline describing the car path along the whole knot, so I could for 
example make an animation, but this would require to read manually the 
position of a few hundreds vertices in wings :-/ I could not figure an 
easy/lazy way to get this info yet...



Thibaut

Post a reply to this message

This is really nice. Took me a couple of seconds to 'see' what I was looking
at! I sympathise with the background problem - I too had trouble thinking of
anything simple and effective. I'm now spending longer on the background
than I am on the knots...

Suggestion, to be used/recoiled from as you see fit: Put a layer of cloud
beneath the knot, then a moonlit landscape (fields, etc) below that. The
effect would then be of the knot floating 1000s of ft in the air, with the
viewpoint even higher but looking straight down.

> The cars are simply placed in pov, using position and angle infos read
> in Wings3D. I just chose the points where I wanted the cars to be, read
> the corresponding position (and angles) in Wings, and then
> translate/rotate the car object as needed in pov.
Another option would be to use trace().

> I wish a had the
> spline describing the car path along the whole knot, so I could for
> example make an animation, but this would require to read manually the
> position of a few hundreds vertices in wings :-/ I could not figure an
> easy/lazy way to get this info yet...
I specifically looked for parametric functions describing knots so I could
generate them algorithmically. If there's some way of persuading knotplot
to give up a mathematical function, it'd be no problem. It might be worth
contacting the author in this regard...

It'd be great to see some more of this work - keep it up!
Bill

Post a reply to this message

Thibaut Jonckheere wrote:


> translate/rotate the car object as needed in pov. I wish a had the 
> spline describing the car path along the whole knot, so I could for 
> example make an animation, but this would require to read manually the 
> position of a few hundreds vertices in wings :-/ I could not figure an 
> easy/lazy way to get this info yet...
> 
Not sure if it is easy/lazy enough but I sometimes make use of the 
ability in Wings to assign materials.  When selected polys are then 
exported into .obj format, the material assignment can be used as a sort 
of label to identify vertices in the .obj file.  I then parse the file 
with an external script ( I like Python ) which puts the vertex/normal 
info into sdl format, arrays for example.

Post a reply to this message

Bill Pragnell wrote:
> This is really nice. Took me a couple of seconds to 'see' what I was looking
> at! 

This couple of seconds interests me ! Can you say if it is because of a 
"technical"problem (image too dark, not enough contrast,...), or because 
it is due to the complexity of the object shown ? I guess many people 
(including me) won't often make the effort, in usual life, to look 
longer at a picture which seems unclear or confuse at  first sight.

I sympathise with the background problem - I too had trouble thinking of
> anything simple and effective. I'm now spending longer on the background
> than I am on the knots...
> 
> Suggestion, to be used/recoiled from as you see fit: Put a layer of cloud
> beneath the knot, then a moonlit landscape (fields, etc) below that. The
> effect would then be of the knot floating 1000s of ft in the air, with the
> viewpoint even higher but looking straight down.

Thanks, I will try something in this vein for the background. I am quite 
conviced now that what is lacking is a feeling of depth behind the knot.


> 
>>The cars are simply placed in pov, using position and angle infos read
>>in Wings3D. I just chose the points where I wanted the cars to be, read
>>the corresponding position (and angles) in Wings, and then
>>translate/rotate the car object as needed in pov.
> 
> Another option would be to use trace().

In fact, I use both: I perform a trace to get the angle between the 
local normal and the vertical (y axis), and the wings info to get the 
rotation angle in the horizontal plane.


> I specifically looked for parametric functions describing knots so I could
> generate them algorithmically. If there's some way of persuading knotplot
> to give up a mathematical function, it'd be no problem. It might be worth
> contacting the author in this regard...

I may have a look at the math formulae behind this knot. But it is too 
late for my image now: my knot does not follow exactly the one given by 
knotplot.

> It'd be great to see some more of this work - keep it up!
> Bill
>

Thanks for the support !


Thibaut

Post a reply to this message

Jim Charter wrote:
> Thibaut Jonckheere wrote:
> 
> 
>> translate/rotate the car object as needed in pov. I wish a had the 
>> spline describing the car path along the whole knot, so I could for 
>> example make an animation, but this would require to read manually the 
>> position of a few hundreds vertices in wings :-/ I could not figure an 
>> easy/lazy way to get this info yet...
>>
> Not sure if it is easy/lazy enough but I sometimes make use of the 
> ability in Wings to assign materials.  When selected polys are then 
> exported into .obj format, the material assignment can be used as a sort 
> of label to identify vertices in the .obj file.  I then parse the file 
> with an external script ( I like Python ) which puts the vertex/normal 
> info into sdl format, arrays for example.


Hum, parsing the .obj export sounds interesting, even if it not totally 
lazy :-)  The car trajectory is a single edge loop in my Wings object, 
so I guess I can bevel it, assign a material to the faces, and export.

Thanks for the tip !


Thibaut

Post a reply to this message

Thibaut Jonckheere wrote:
> it is due to the complexity of the object shown ? 

In my case, a lack of clues about the scale. It wasn't obvious 
immediately that the little dots of lights were automobiles.

Personally, I like that little delay. It's what art is about.

-- 
   Darren New / San Diego, CA, USA (PST)
     "I think these anchovies are spoiled.
      They're not flat."

Post a reply to this message

Thibaut Jonckheere <tua### [at] yahoofr> wrote:
>I wish a had the
> spline describing the car path along the whole knot, so I could for
> example make an animation, but this would require to read manually the
> position of a few hundreds vertices in wings :-/ I could not figure an
> easy/lazy way to get this info yet...
>
>
>
> Thibaut

You might check out my knotsplines macro, where I describe how to get a
spline from the povray-exported bicubic patch done by KnotPlot.

Basically, I use Microsoft Word (Python or something similar would be much
easier, I'm guessing), to take the first vertex point from each patch (I
know, it would be more accurate to take one of the interior ones, since the
"edge" vertices are control points....), and put them into an array, and
form the spline from the array -- the readme/macro shows how:

http://tinyurl.com/lfvar

(although, when I try to save the zip file over the web view, it comes
through corrupted; when I try to download it with the newsreader it works
fine.  I'm not sure why.)

If you'd like a copy, I can send you one.

Dave Matthews

Post a reply to this message

"Bill Pragnell" <bil### [at] hotmailcom> wrote:
> If there's some way of persuading knotplot
> to give up a mathematical function, it'd be no problem. It might be worth
> contacting the author in this regard...
>

I don't think that KnotPlot ever figures an explicit parametric
representation.  I think it works from points (the "beads") and splines.
That's a frustration for those of us more used to calculus and closed-form
solutions; but, I think, once one gets used to approximating from spline
points, the calculations can be much quicker and simpler (I'm still working
out a few of these, when I get the time, which is scarce right now....)
Also, I'd imagine that the parametric equations for some of the more
complicated knots would be hideous.

Dave Matthews

Post a reply to this message