looking for some feed back on a date palm macro I'm working on.

-all csg with a sphere sweep for the trunk.
-manual bounding of branches and leaves gave a 350% speed increase.
-short stubby to tall and thin trees can be made.
-change random seed for a complete new tree.
-random noise in scaling and 3 axis rotation of each leaf gives nice realism
close up.
-texture is added to each branch during placement and changes depending upon
elevation. - they get older the lower they are.
-used a tan() in vertical rotation so majority of branches are out to the side
and fewer point up or down.
-various randomness throughout for realism (I hope)
for example in branch placement:
    rotate (.5-rand(Rn))*B*z
    rotate <tan(1.16-Sx*2.32)*39+5, 0, (.5-rand(Rn))*20>
    translate <0, (Sx-.4)*1.25+H+.4, min(1,1.4-Sx)*W/.8>
    rotate rand(Rn)*360*y

trees as rendered called with below settings

// scale, rotate, trans, height, trnk width, trnk crv amt, num of palms, palm
twist base, palm twist increment, dead hangers, num of dates, rnd seed

palmtree1(1,<0, 50, 0>,<48, -.3, 80>,30,.5,1.5,70,25,5,7,14,24)
palmtree1(.8,<0, -170, 0>,<43, -.3, 84>,29,.6,5,55,30,6,0,10,410)

image is just a small clipping of the tree from a larger image.
med. radiosity and AA so quality can go way up.
will be adding SSLT support for the textures and custom textures.
will post macro with instructions later.

feed back please, good and bad, thanks.

Post a reply to this message

Attachments:
Download 'palm.png' (170 KB)

Preview of image 'palm.png'

Looking good so far, although I would prefer a close up on the tree 
and/or details to really be able to comment.

Thomas

Post a reply to this message

"OJD" <nomail@nomail> wrote:
> looking for some feed back on a date palm macro I'm working on.
>...
> -manual bounding of branches and leaves gave a 350% speed increase.

Wow! That's interesting to know. I don't understand how or why it works, though.
I would assume that auto-bounding would give the best speed(?) unless there are
difference/merge CSG objects present.

>
> trees as rendered called with below settings
>
> // scale, rotate, trans, height, trnk width, trnk crv amt, num of palms, palm
> twist base, palm twist increment, dead hangers, num of dates, rnd seed

A *long* macro entry! But obviously necessary. I have a model of my own --my
'generic human' CSG model --that I was thinking of recoding for animation, as a
macro. It would probably require this many macro entries and more, for all the
joint rotations. I had not attempted it for just that reason-- as seemingly a
rather cumbersome macro to use -- but it's nice to know that someone else is
using more than just a few parameters, when necessary. So I guess a lengthy
macro is not such a foreign concept after all. Your set-up inspires me. :-)

I really like your image, the palm fronds look quite nice and realistic. I'd
love to see a close-up.

Post a reply to this message

On Fri, 17 May 2013 22:26:54 +0200, Kenneth <kdw### [at] gmailcom> wrote:


> So I guess a lengthy macro is not such a foreign concept after all.:-)

And I thought I was the only one...
Drips (Obj, ColMap, CamPos, CamLook, CamDirect, image_width, image_height,  
ColVar, Finish, Spacing, MaxDrops, MaxLength, MinThickness, MaxThickness,  
Merge, Seed, Squiggliness, ColSystem, Threshold, Direction,  
ContainerStart, ContainerEnd )

;)

-- 
-Nekar Xenos-

Post a reply to this message

"OJD" <nomail@nomail> wrote:
> -random noise in scaling and 3 axis rotation of each leaf gives nice realism
> close up.

I would like to see a higher res image but at this scale it looks great.

Post a reply to this message

"Kenneth" <kdw### [at] gmailcom> wrote:
> "OJD" <nomail@nomail> wrote:
> > looking for some feed back on a date palm macro I'm working on.
> >...
> > -manual bounding of branches and leaves gave a 350% speed increase.
>
> Wow! That's interesting to know. I don't understand how or why it works, though.
> I would assume that auto-bounding would give the best speed(?) unless there are
> difference/merge CSG objects present.
>

the leaf is made from a difference of a sphere and 2 cylinders giving a curved
slice of the top of the sphere. then it is differenced from itself scaled
slightly larger. this gives a nice curved 'V' shaped leaf.

now the thin slice left after the initial difference is much smaller than the
auto bounding boxes created as they want to cover all the objects in the
difference. compound this with using the shape many times in close proximity
causes many overlapping bounding boxes with the majority of the intersection
tests failing (only about .04% succed)]

after manual bounding, the intersection test success ratio goes way up with a
much smaller number of tests performed.



----------

here is a test image from earlier stage of development.
 the twisting is a bit radical... lol
 many changes were made since this render:
- textures are better
- branch tips are better. filled in the empty v shape.
- leaf was redone
- elevations of branches are much more realistic by adding the tan()
- branches near the top are shorter than the side branches

but this gives you a better view of the thing.

renderer is busy with a long test but will get an updated version rendered in a
day or so.

Post a reply to this message

Attachments:
Download 'palm2.png' (676 KB)

Preview of image 'palm2.png'

hear is a low quality render early in development showing a close up of the leaf
shape.

 noise in the rotation of each leaf adds a lot to simulate real world palms.
     rotate (rand(Rn)*1.2-5-XX)*9*y
     rotate ((XX+4)*I+B+((rand(Rn)-.4)*4-XX*5))*z

 branch is actually 2 of the leaf sections added end to end without the v
cutout.

 leafs follow the branch with a sin() function
    translate <-.01, sin(pi/2-XX/5.6)-.03, XX>



.... ignore the stuff behind the leafs... lol

Post a reply to this message

Attachments:
Download 'palmleaf.png' (219 KB)

Preview of image 'palmleaf.png'

The tree is looking really good indeed. Excellent work.

Do you know the palm tree made by POV-Tree (Gena Obukov's java version 
of Tom-tree, by Tom Aust)? This is how it looks as CSG output.

Thomas

Post a reply to this message

Attachments:
Download 'povtree_test.png' (188 KB)

Preview of image 'povtree_test.png'

"Nekar Xenos" <nek### [at] gmailcom> wrote:
> On Fri, 17 May 2013 22:26:54 +0200, Kenneth <kdw### [at] gmailcom> wrote:
>
>
> > So I guess a lengthy macro is not such a foreign concept after all.:-)
>
> And I thought I was the only one...
> Drips (Obj, ColMap, CamPos, CamLook, CamDirect, image_width, image_height,
> ColVar, Finish, Spacing, MaxDrops, MaxLength, MinThickness, MaxThickness,
> Merge, Seed, Squiggliness, ColSystem, Threshold, Direction,
> ContainerStart, ContainerEnd )
>
> ;)
>
Impressive! ;-)  I had a funny feeling that your macro might be like this.

Post a reply to this message

"OJD" <nomail@nomail> wrote:

>
> now the thin slice left after the initial difference is much smaller than the
> auto bounding boxes created as they want to cover all the objects in the
> difference. compound this with using the shape many times in close proximity
> causes many overlapping bounding boxes with the majority of the intersection
> tests failing (only about .04% succed)]
>
 Ah, of course; that makes sense. For some goofy reason I was thinking your
fronds were simple mesh objects--I should have paid closer attention to the very
first entry in your original post:  CSG

Post a reply to this message