varol@cwi.nl (Varol Akman) (03/08/88)
I am looking for a large (i.e. more than 1000000 edges) graphics object. I'll use it for non-profit research purposes--to test an algorithm. I would be more interested in public domain objects but anything else is also okay. The format of the object is no problem as long as it can be converted (with little or medium) effort to a Brep. Please reply me personally so that we don't clutter the network. Thank you. -Varol Akman Center for Math. and CS Amsterdam, the Netherlands
blob@calgary.UUCP (Brian Wyvill) (03/10/88)
In article <235@piring.cwi.nl>, varol@cwi.nl (Varol Akman) writes: > I am looking for a large (i.e. more than 1000000 edges) graphics > object. I'll use it for non-profit research purposes--to test an > algorithm. A lot of people are doing research which involves testing algorithms (such as ray tracers) with large objects. I have seen several papers where the authors use objects of only 4 or 5 thousand polygons. In practice, scenes often require hundreds of thousands of polygons and the use of standard objects would help in comparing algorithms. Sending test objects of this size through the net may not be practical. One solution is to use objects which can be generated in a standard fashion. For example the Menger sponge (See Mandelbrot) or the recursive tetrahedrons which have been used by some researchers (see proc. SIGGRAPH) Such objects can be manufactured with a recursive procedure and at different recursion levels provides discrete numbers of polygons, for example the Menger Sponge increases by a factor of 20 at every recursion level. I have a definition of the Menger Sponge written in the graphics language: PG (Polygon Groper), if there is any interest I will post the program, it may be of some use even without the PG interpretter. -- Brian Wyvill ..!{ubc-vision,ihnp4}!alberta!calgary!blob
eric@batcomputer.tn.cornell.edu (Eric Fielding) (03/11/88)
In a recent article blob@calgary.UUCP (Brian Wyvill) wrote: >In article <235@piring.cwi.nl>, varol@cwi.nl (Varol Akman) writes: >> I am looking for a large (i.e. more than 1000000 edges) graphics >> object. I'll use it for non-profit research purposes--to test an >> algorithm. >A lot of people are doing research which involves testing algorithms (such >as ray tracers) with large objects. I have seen several papers where the >authors use objects of only 4 or 5 thousand polygons. In practice, scenes >often require hundreds of thousands of polygons and the use of standard >objects would help in comparing algorithms. [...] > Brian Wyvill I have been torture testing a local ray-tracing program with some *real* data. It generallyu uses up all of the virtual memory that we have at about 80,000 triangles. The data is the topography of the Central Andes in South America, and the full data set is about 100,000 grid points, which converts to 200,000 triangles. I have another topographic data set that is 1200x3600 elevations. Rendering this would require > 8 million triangles. It would be fairly compact to mail, since one only needs the array of values to generate the surface. Of course, I would love to see a Gouraud (sp?) shaded version ;-). Has anyone developed algorithms that take advantage of the assumptions one can make about a surface that is basically a large rectangle with fairly minor perturbations on it? Seems like it could be rendered much more efficently by not just treating it as millions of triangles. If anyone wants some topography to render, let me know. I would like to see any results, if possible. I can also provide my code for turning it into colored triangles. ++Eric Fielding eric@tcgould.tn.cornell.edu eric@crnlthry.bitnet !cornell!batcomputer!eric