BIO My name is Pavel Fadrhonc and I'm first year master student of Faculty of Information Technology in Brno, Czech Republic and my major is Computer Graphics.
I have 3 year experience on c++ programming and 1 year of OpenGL. Last year I finished bachelor study with thesis on theme of '3D Landscape Rendering with Lightmap' implemented in c++ and OGL. Last semester I worked in 3-members team on project of Volume Rendering. Other than that I made bunch of other programming projects for school althougt not related to computer graphics, like BSD socket programm for P2P chatting for which I designed and implemented my own net protocol. I'm also working as programmer in Czech telephone company IPEX, currently developing web frontend in PHP/Javascript as part-time job. CRYSTALSPACE QUESTIONS: My last exam is 11.6 so I can't promise fulltime working until then but this will be compensated by more work during actual coding period since my next semester starts by beginning of October. I don't have any other plans during summer coding period. I might be applying for OGRE and Battle for Wesnoth this year but I'm not sure if will have some time for it. As regarded to my Python experience, I had course about scripting languages last year and I made project on dowloading and aligning english and czech subtitles for it in Python. Last year I was applying to CrystalSpace on idea of Global illumination using photon mapping but I wasn't selected. This year I propose implementing of Automating managing of Impostors. IDEA: Currently, CrystalSpace support regenerating impostors based on distance and rotation of camera and object position regarding to the center of the screen. This is fine and basic solution and I propose to extend this solution by implementing of algorithm found in this paper: http://www.cg.tuwien.ac.at/research/publications/2005/jeschke-05-AIP/jeschke-05-AIP-Paper.pdf This system takes care of generating and automatic regerating of impostors in relation to given area of view a image quality while preserving low momery requirements and, of course, desired target minimum framerate. Main atributes of this algorithm are: - acceleration techiques like visibility culling and level of detail are applied before rendering impostors - if there are more object in adjacent space, one impostor is generated for all of them to save memory - using of shaft view cell instead of classic rectangular view cell. Shafts offer minimum distance and user defined angle. This technique is already aplemented in CrystalSpace as SetRotationTolerance virtual method for impostor, so it will be used in final product. Algorithm in nutshell: 1) Objects are clustered using KDtree technique already implemented in CrystalSpace. 2) Problematic views are found and hierachically subdivided to fulfil user defined accuracy. 3) Impostor candidates are generated. Here, impostors are merged while in adjacent space and shaft view cell is used so the impostor require less memory. This is due the fact, that impostor is generated for whole one region from that part of scene that impostor represents and that's appearance changes a little from given distance. 4) In this step, optimizations are in place. Best candidates for each problematic view(i.e. view that suffer from framerate drop due to big complexity of that part of scene that impostor represents) are selected. TIMELINE Timeline comes as soon as I get some reactions on this, first version of my proposal. I want to know, what you guys think about my idea. I'm really looking forward to work with CrystalSpace, since I found it very exciting and interesting project. I love computer graphics programming and in the future, it is my long-term goal to work as computer graphics programmer, preferably, in c++. As I was checking out projects code, I really liked the fact, that CrystalSpace utilized most if not all of C++ advanced features, as I would like to work on such subtle and sophisticated project. ------------------------------------------------------------------------------ _______________________________________________ Crystal-main mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/crystal-main Unsubscribe: mailto:[email protected]?subject=unsubscribe
