Scene Graphs?

I want to use a Scene Graph in my application.
Which is the best one? (on performance)

I’m thinking about using the OpenSG, have you tried it?

Originally posted by aleixventa:
I want to use a Scene Graph in my application.
Which is the best one? (on performance)

There are plenty of both commerical and open source scene graphs out there, which is best for your purpose depends upon the focus of your app.

A selection of the commerical scene graphs

Gizmo3D
Inventor
OpenGVS
Performer
Vega Prime
VTree

A selection of the open source scene graphs (in alpabetic order).

OpenRM
OpenSceneGraph
OpenSG
PLIB
SGL

Lots to chose from, an indication of good a scene graph is likely to be related to how many users have made the decision to use it, which in turn will related who active the support mailing lists will be. So I’ve totalled up the previous three months mailing list activity (Sep,Oct,Nov) :

OpenRM 12
OpenSceneGraph 1590
OpenSG 367
PLIB 655
SGL 78

As percentages, and in order

OpenRM 0.4%
SGL 3%
OpenSG 14%
PLIB 24%
OpenSceneGraph 59%

I have to admit, I wouldn’t have posted these figures if they I’d didn’t think the the OpenSceneGraph would come near the top, but you can check the archives yourself nothing has been biased in anyway.

As for trends, OpenRM and SGL are very small. OpenSG is small, but growing. PLib is a reasonable size but activity looks like its leveled off. This time last year OpenSceneGraph was smaller than Plib, has grown three fold in that period.

This time next year it looks like OpenSG might over take Plib, but its probably too specialized and complex for really wide spread adoption. Who knows how big OpenSceneGraph will be next year, we’ll be at 1.0, the books will be out or on there way…

I’m afriad I can’t provide any benchmark comparison as unforntately there is little overlap between the file formats all the projects can load, plus I’m really to busy to get them all to compile and tested…

Tests between the OpenSceneGraph and the commericial scene graphs show that its on par and sometimes better.

You have to do tests yourself, on your own data, and take a hard look at the features that you need.

Good luck,
Robert