Most efficient way to model thousands of trees?

[Christiaan]

We commissioned a large city wide context model from a third party and it includes roughly 3000 trees. The trees in the model are simplified lollipop types, with each tree in the model being a scaled version of one symbol. The symbol itself is made up of simple 3D polygons (as attached).

 

It works, but I'm wondering if I can make it more efficient. I've changed the trunk to an extruded circle. But what about the canopy? 

 

I was thinking a Sweep or a Subdivision? I tried the Subdivision but it doesn't appear to be compatible with scaled symbols. I also tried a sweep and while this reduced the file size by a factor of about 60 the problem is when I go into 3D view (of 3000 of them) it swallows up all my free memory (about 15 GBs of it) and brings everything to a screeching halt. Which it doesn't do with the 3D polygon version.

 

 

Tree-3D-Polygons.vwx

Tree-Extrude-Sweep.vwx

[Kevin McAllister]

55 minutes ago, Christiaan said:

I also tried a sweep and while this reduced the file size by a factor of about 60 the problem is when I go into 3D view (of 3000 of them) it swallows up all my free memory (about 15 GBs of it) and brings everything to a screeching halt. Which it doesn't do with the 3D polygon version.

 

Increase the segment angle of your sweep. Right now its set at a tiny 0.70 degrees. I experimented a bit and I think you could go to 20-25 degrees without a visible difference.

 

Kevin

 

Tree-Extrude-Sweep_20 degree.vwx

Edited November 6, 2017 by Kevin McAllister

[Christiaan]

Thanks Kevin, that's the ticket. I went with 36º, which is fine for our purposes.

 

I also changed 3D conversion under 3D Preferences to medium. I think that also affects things in this regard, right?

[Matt Overton]

Have you considered just making the trees 2-3 crossed planes.

[Benson Shaw]

Just a couple comments in addition to reducing the symbol vertex count for more efficient navigation. Other efficiencies might be in order

1. General:

•  Nice work on the original drawing! That’s a lot of trees, a lot of work. Navigation issues aside, the drawing looks great! I hope to see further landscape and architecture features.
•  In a duplicate vwx file, I replaced all the trees with Kevin’s modified symbol via Delete Symbol>Replace with . . .  Confirming that Navigation is much better with the simpler symbol. 

2.  Symbol Components
Christian’s SKP LollyPop tree symbol is a nest containing:

a. Tree_DWG symbol which contains:

b. Group2 (Canopy) symbol and the Group3 (Trunk) symbol.

c. Would the drawing navigation be more efficient by eliminating the outer SKP container? (see below)

d. Simpler overall symbol ala @Matt Overtonmight be worth looking at.

 

3. Symbol Scale
a. The unscaled Tree_DWG symbol is 247m tall (Meters!) with the canopy about 3/4 of the height. That’s a big tree!? Why???
b. Tree_DWG symbol is scaled by .0187 inside the SKPLollyPop container symbol.
c. which makes the unscaled LollyPop 4.625 meters tall, by 2.391 meters diam.
d. which makes the various scales of the 2700+ instances in the drawing, um, awkward. OIP scale values for any of the instances are uninformative.

 

4. Alternate symbol scale idea
a. Create a “unit” symbol that has native scale of 1m tall by 1m diam.  Insertion point is center of trunk footprint.
b. Too late to redo the scale of 2700+ symbols on this drawing, but for future work wouldn’t it be more efficient to place & scale this “unit” dimensioned symbol?
c.  Scaling a unit symbol to match survey data would be easy. Example, Survey says 9.6 meters tall x 3 meters diam, the OIP symbol scale factors are x=3, y=3.  z = 9.6.
d. The scale factor values in OIP indicate the tree dimensions without further conversion.
e. Could a Marionette network read the survey data (location xyz plus tree height/diameter) and automatically place and scale each symbol instance?

 

5.Notes:
a. The trees in the drawing have widely varying bottom z values, often within a meter or 2 of z=0, presumably responding to terrain.  But several are way up there at 14m or so.  Top of a steep berm? Might be worth checking.
c. I tried a mildly complex lofted symbol to test navigation lags with lofts vs sweeps. About the same.
d. The vwx file takes a LONG time to launch, whether orig symbols or Kevin’s or my own tests. The Verify step seems endless, about 3.5 min.

 

OK, enough!

-B

[Kevin McAllister]

1 hour ago, Benson Shaw said:

OK, enough!

-B

Great breakdown Benson! I'm really missing the reputation buttons right now 

Kevin

[Christiaan]

12 hours ago, Matt Overton said:

Have you considered just making the trees 2-3 crossed planes.

 

I hadn't, because we quite like the shading in top view. But I'll explore it, thanks.

 

 

11 hours ago, Benson Shaw said:

Nice work on the original drawing! That’s a lot of trees, a lot of work. Navigation issues aside, the drawing looks great! I hope to see further landscape and architecture features.

 

Oops, I only meant to send one tree, not all of them . Thanks for your great feedback Benson, much appreciated. Although we didn't model the trees, more on that below. Here's a view of the whole model:

 

 

11 hours ago, Benson Shaw said:

The unscaled Tree_DWG symbol is 247m tall (Meters!) with the canopy about 3/4 of the height. That’s a big tree!? Why???

 

This is how we received it. It comes from a company that uses high-res imagery to create these models. Initially we received the model as a DWG file, and the trees came as individual ungrouped 3D polygons, which wasn't very useful. When we asked for the trees as individual objects that came back with a solution whereby they ran their model through Sketchup. And this is what we received. So, yes, a bit messy, but it did the trick.

 

11 hours ago, Benson Shaw said:

Too late to redo the scale of 2700+ symbols on this drawing

 

You think so? I'd like to try. My machine may choke though.

 

11 hours ago, Benson Shaw said:

The trees in the drawing have widely varying bottom z values, often within a meter or 2 of z=0, presumably responding to terrain.  But several are way up there at 14m or so.  Top of a steep berm? Might be worth checking.

 

 

Yes, all following terrain correctly. 

 

Thanks again Benson. Super feedback

 

Edited November 8, 2017 by Christiaan

[Benson Shaw]

Wowzer! What a cool project!  Thanks for letting us see your fine model of this expansive project. Looks like a boatload of work ahead.

-B

[rowbear97]

You could make the cross plane trees a hybrid object with the tree canopy showing only in top/plan view and use the drop shadow setting to get the desired shade effect.

[AlanW]

@ChristiaanThese Marionette items create a series of plant symbols at various densities. Simply change the polygon shape and the scale and density. Create your own symbols as needed.

HTH

Just noticed that in 2018 these files do not read the correct symbols.Maybe try in 2017 or maybe your file wil be ok. All used to work well when i made them back when?

 

Will look to see why they dont operate correctly now.

 

Ok not sure why but on the tree Marionette object, enter the network and then exit and all seems to read the correct tree symbol.

 

Random_Planting_001_v2018.vwx

Random_Planting_IProp_001v2018.vwx

Edited November 27, 2017 by Alan Woodwell