Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Wed Oct 10, 2007 2:49 pm Post subject:
Night Builds news coverage.
Subject description: Auto normals is getting better.
Hello everyone! Right now, I'm in the computer graphics conference where I've presented the Tangent Plane Auto Normals technique seen in VXLSE III 1.37. I'll bring some photos later, since my only internet access here is through the wi-fi from the conference and today is the last day here and.. I'm about to leave. Anyway, I've recently uploaded a new alpha version of VXLSE III with even better auto normals. And, no, they are not perfect. But the results are already interesting and much better than the one from 1.37.
So, here's what we have so far in our SVN:
-> VXLSE III 1.38 Alpha, Revision 17: Compared with 1.37, this version is stable and, likewise 1.37 it has problems when you try to view the whole model in the 3D preview window. The difference is really the quality of the auto normals, which also comes with a work in progress feature known as 'Treat Descontinuities' (quite technical name, I know) which changes the way it finds the normals. It is a bit more acurate, but slower.
-> OS BIG Editor 0.58 Beta, Revision 41: Compared with 0.5 Beta, this version comes with a lot of new features. As long as you don't rename files, this version seems to be pretty much stable. Renaming files might cause corruption. Anyway, it displays your files in Tree View, like windows explorer. It allows you to save .BIG files with and without refpack compression, options are now in an ini file, it has multi-language support under construction, it allows you to extract directories, you can also drag files from windows explorer to drop into it, add directories and much more.
The OS SHP Builder there is the same beta version available in the beta forums, but now guests can also grab them.
So, for those wondering how to download these files, head to the recently updated PPM FAQ to find out. If you want to contribute with the development of these tools, you can report bugs of these programs on their respective forums or even contact me to get authorization to modify these files if you wish to do so. If you want me to shift my focus to a tool that interests you more, the donate button is in the right side and PM me or email me with what you want to be done. As a capitalist pig, I'll follow the best offer, of course. QUICK_EDIT
Nice work Banshee. _________________ Micro TS
Portable, no campaign, movies or music, just the engine and needed resources for skirmish and lan. QUICK_EDIT
Hey Banshee, is there any way of accessing your article online (for 'free')? The answer is most likely obvious, but worth posting if anyone else has the same question. We could peer review it for you so to speak, even though many other people have probably already done so (such as your professor / lecturer) - just out of interest. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Thu Oct 11, 2007 8:20 pm Post subject:
Equiredox wrote:
Hey Banshee, is there any way of accessing your article online (for 'free')? The answer is most likely obvious, but worth posting if anyone else has the same question. We could peer review it for you so to speak, even though many other people have probably already done so (such as your professor / lecturer) - just out of interest.
The article was writen in english because visitors of PPM would be able to read it later. So, I'll host it here once I return from Minas. Right now my internet conection is horrible, but better than nothing. QUICK_EDIT
The article was writen in english because visitors of PPM would be able to read it later. So, I'll host it here once I return from Minas. Right now my internet conection is horrible, but better than nothing.
I suppose your article will answer all these questions, if not, it would be great if you could answer or elaborate on them:
*What is the major advantage of using your technique over the current accepted auto-normal algorithms?
*Does it have better space and or time complexities on average or worst case then the current auto-normal algorithms?
*Better quality output then the current auto-normal algorithms?
*Adaptive to a multitude of input data structure types?
*What family of algorithm does it belong to (D&C etc.)?
These are just the sort of questions I have. No offense, sounds cool - and you must have done something awesome to attend such an event. Just want you to elaborate on what you did awesome with your algorithm. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Oct 12, 2007 12:58 pm Post subject:
Quote:
*What is the major advantage of using your technique over the current accepted auto-normal algorithms?
Most of auto-normal algorithms I've seen in papers are usually a simplified thing and the papers that have them have other purposes.
I've also seen a paper that proposes to find normals by using templates. The purpose of this algorithm is to find any normal, not by using templates.
At the conference, a reviewer told me that there was a somewhat similar technique described at Using Distance Maps for Accurate Surface Representation in Sampled Volumes, by Sarah F. F. Gibson from Mitsubishi, but I never found that before I went to the conference.
Quote:
*Does it have better space and or time complexities on average or worst case then the current auto-normal algorithms?
No, it's a bit heavier than most existing auto-normals algorithm, because it focus on quality results, although it is still fast (O(m³n), where m is the range (usually a small number) and n is the number of voxels.
Quote:
*Better quality output then the current auto-normal algorithms?
Theorically yes. My implementation still has some issues that needs to be sorted.
Quote:
*Adaptive to a multitude of input data structure types?
No, it is restricted to voxels being faces and only reflecting light to only one direction, although it isn't hard to modify it for double facing. Actually, you just need to remove the ray casting procedure in the end and give the two normals.
Quote:
*What family of algorithm does it belong to (D&C etc.)?
No idea. I did the whole thing from scratch, to be honest. QUICK_EDIT
Also Known As: banshee_revora (Steam) Joined: 15 Aug 2002 Location: Brazil
Posted: Fri Oct 12, 2007 6:48 pm Post subject:
Update: Another revision of VXLSE III has been uploaded in order to improve the auto normals. It's really under research stage and I'm testing a lot of things. So far, this is the best setting I could find... QUICK_EDIT
You cannot post new topics in this forum You can reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum You cannot attach files in this forum You can download files in this forum