PDA

View Full Version : has this been done before?



zed
11-18-2002, 09:49 PM
gidday ive maybe invented a technique which does softshadows (it also selfshadows + also handles the bumpmaps, ie each bump casts a shadow)

a quick picture, which u might wanna turn up contrast or something to see the shadow, i still havent worked out the finaer points (time is short at the moment)
http://uk.geocities.com/sloppyturds/t5.jpg
(theres no t1.jpg etc there so dont bother looking http://www.opengl.org/discussion_boards/ubb/smile.gif)

now does anyone know that this sort of thing has been done before as in the past ive thought ive invented things (only to find out later some mother had done then 20 years before http://www.opengl.org/discussion_boards/ubb/smile.gif)

for the curious
*i dont think this can be done in hardware on a gf2, 85% of the equation can, but ill think a bit more about it so who knows.
*but i see no reason why it cant be done in hardware on more powerful cards (note i dont know exactly what these more powerful cards can do fully so take this with a grain of salt http://www.opengl.org/discussion_boards/ubb/smile.gif )
*also unfortunatly this aint radiosity (im still working on that)

davepermen
11-18-2002, 10:03 PM
well, softshadows where done before, with bumps that do shadows, too..
important is _how_ you do it.. did you told us?

and about the newer cards.. we don't know eighter if you don't tell us how http://www.opengl.org/discussion_boards/ubb/biggrin.gif (if you don't want talk open about it, you can drop a mail, you can trust me.. hehe http://www.opengl.org/discussion_boards/ubb/biggrin.gif)

if 85% fits into gf2, i'm pretty sure it works on gl1.5 hw.. dunno, possibly even g3/gf4..

[This message has been edited by davepermen (edited 11-19-2002).]

SergeVrt
11-19-2002, 01:08 AM
Originally posted by zed:

http://uk.geocities.com/sloppyturds/t5.jpg
(theres no t1.jpg etc there so dont bother looking http://www.opengl.org/discussion_boards/ubb/smile.gif)




Heh, Yahoo says on you link .....
-->
We're sorry, but this page is currently unavailable for viewing. Please try again later, or search Yahoo! GeoCities for other pages that interest you.

davepermen
11-19-2002, 01:17 AM
Originally posted by SergeVrt:

Heh, Yahoo says on you link .....
-->
We're sorry, but this page is currently unavailable for viewing. Please try again later, or search Yahoo! GeoCities for other pages that interest you.



copy and paste the url, then it works.. and for the stupid ones, i uploaded it here..
http://www.itstudents.ch/users/dave/free/files/t5.jpg

cbwan
11-19-2002, 02:02 AM
fyi it failed to work when I clicked,
then I tried copy/paste, still didn't work.
and 2 mins later everything worked fine.

I guess I must twice as stupid http://www.opengl.org/discussion_boards/ubb/wink.gif

so, has this been done before or not?
what is the technique?

cb
--
“Stupid is who stupid does”
-- forrest gump

Matt Halpin
11-19-2002, 02:23 AM
tbh, you'll have to tell us your technique before we can say. I'd be surprised if it's completely new though. http://www.opengl.org/discussion_boards/ubb/smile.gif

After all, there are quite a lot of people experimenting with soft shadows (including me). It's one of the most obvious areas to look at.

I'm using the fairly standard technique of blurring the shadow mask and using pixel depth to choose the amount of blur...

Ysaneya
11-19-2002, 02:43 AM
That's also what i'm basically doing, except that i calculate the distance to the camera and to the light per vertex.

Some screenshots: http://www.fl-tw.com/opengl/SoftShadows/

Y.

Matt Halpin
11-19-2002, 03:17 AM
Ysaneya:

How are you coping with shadow bleeding?

ie, when an object that is in light is in front of an object that is in shadow. The blurring means that the light bleeds over the shadowed object in the background. Looks weird. Any ideas?

On better hardware you can do a weighted blur where each sample is weighted based on it's depth difference from the centre sample point. This means that blurring doesn't pass over object edges. I'm on a GF3 though. http://www.opengl.org/discussion_boards/ubb/smile.gif

Ysaneya
11-19-2002, 05:04 AM
The amount of blurring is based on the distance to the camera, so theorically, if you've got a big object in front of you that receives a shadow, and a far-away background, the artifacts are minimized (i think you can see that in the first screenshot: the shadows on the trees are "sharp". Yet these artifacts are still very visible in some conditions.. a dark ambiant color makes things even worse. Unfortunately i don't think there's a magical solution, i just try to keep things as good as possible :)

Y.

davepermen
11-19-2002, 05:08 AM
the reflections don't have shadows, right? i don't see the shadow in the sea..

Ysaneya
11-19-2002, 05:54 AM
No; actually they don't even have lighting (not in these shots, i've fixed it in a more recent version). When it's animated you don't notice shadows anyway.

Y.

davepermen
11-19-2002, 06:03 AM
Originally posted by Ysaneya:
No; actually they don't even have lighting (not in these shots, i've fixed it in a more recent version). When it's animated you don't notice shadows anyway.

Y.

3 words:
when
demo
out
and a fourth:
finally
?

i have radeon9700, want to feed it with stuff http://www.opengl.org/discussion_boards/ubb/biggrin.gif

zed
11-19-2002, 09:09 AM
ive seen a demo where bumps shadow (but only onto the same object, this handles the case where bumps shadow this + other objects)

>>That's also what i'm basically doing, except that i calculate the distance to the camera and to the light per vertex.<<

nice but im not doing per vertex. (i think my technique is completely different)

thinking about my technique more i believe it is original (it requires u to think out of the box 3!!! times)

so my question is
what happens now?
(im not interested in making money from it, patenting it, whatever)
do i write a paper + simit it to siggraph?

Ysaneya
11-19-2002, 09:25 AM
If you're not interested in money/patents, i'd just suggest to share with us, at least you could get some opinions on how good it is :)

Soft shadows do not come without a cost. Is your method fast ? What kind of hardware does it require ( GF3/GF4 only ? ) ? What about the shadows quality ? Is there some artifacts ? From your screenshot it looks like you're not using the distance to the occluder to sharpen it ? I hope you get my point: there's tons of ways to do soft shadows in real-time, but they all have different advantages / drawbacks, and i doubt yours is different :)

Y.

Ysaneya
11-19-2002, 09:30 AM
> 3 words:
> when
> demo
> out
> and a fourth:
> finally
> ?

Hehe, i'm sorry. My 3D artist is busy (you know how these damn artists are! grr), so he hasn't completed the test scene. I wanted it to be nicely textured, with bumpmaps and all the stuff, but it's going slowly. Maybe i should just release what i got so far. Also, for a number of reasons (including: bugged drivers!) i couldn't use vertex shaders as i wanted, so performance has not improved ( i get around 15 fps on my system: 1.4 Ghz athlon, R8500 + 512 Mb RAM ). Finally i broke some stuff by doing more tests, so i need to fix everything.. knowing the number of Radeon users, i was under the impression not a lot of people could run it anyway.. and i lack the time to implement it on a GF3/GF4 :)

Y.

davepermen
11-19-2002, 10:07 AM
heh, always the artists..

to zed, just _show_ that you have something new: write some paper, show a running demo, etc etc etc.. or simply explain the algo to us..

MZ
11-19-2002, 10:29 AM
... or submit it to next Cg contest, at least you'll have chance to get the "hairdryer" (nv30) for free http://www.opengl.org/discussion_boards/ubb/wink.gif

zed
11-19-2002, 10:36 AM
>>there's tons of ways to do soft shadows in real-time<<

true + thats the reason im asking

'does anyone know of any method that does what ive described?'

Ysaneya your method i take it doesnt handle shadows from bumpmaps (thus its a different method)

there are ppl on this board with more 3d knowledge than me eg dorbie,jwatte,cass etc who would surely know if something comparable has been done before (which would save me a lot of time)

ps the reason im not willing to share further info at the moment is cause i have a feeling this is gonna be very big i dont want someone else stealing the credit eg like tesla(or whoever that was)

Ysaneya
11-19-2002, 11:36 AM
No, it doesn't handle bumps shadows. At least if you don't describe it you could say what are its advantages/drawbacks as well as hardware requirements ?

Y.

roffe
11-19-2002, 12:07 PM
ps the reason im not willing to share further info at the moment is cause i have a feeling this is gonna be very big i dont want someone else stealing the credit eg like tesla(or whoever that was)[/B]

Well how accurate, compared to existing shadow generation techniques
and global illumination models, does your solution simulate light object
interaction?

If it is as accurate, or even more, I would say your are on to something truly big.

Humus
11-19-2002, 12:42 PM
Originally posted by MZ:
<snip>the "hairdryer" (nv30)<snip>

LOL, yeah, that was my thought too when I was that cooling solution http://www.opengl.org/discussion_boards/ubb/biggrin.gif


[This message has been edited by Humus (edited 11-19-2002).]

tempyjoe
11-19-2002, 04:19 PM
Hey guys working with soft shadows, don't know if this is of much interest, but Epic showed off a tech demo for GeForceFX (NV30) at COMDEX today.

This is the only screenshot I have seen so far http://gamespot.com/gamespot/stories/news/0,10870,2898008,00.html

However they are claiming that the lighting in that scene is 100% dynamic.

To quote one of the Epic guys

"those are completely dynamic, self-shadowing, soft shadows capable of casting from multiple light sources including ambient environment. "
http://www.shacknews.com/ja.zz?comments=23741&6026615

jpeter
11-19-2002, 04:28 PM
bigest size image http://www.guru3d.com/tech/geforcefx/img/Games/Unreal_GFFX_01.jpg

zed
11-19-2002, 05:12 PM
guys i love thread hijackers and all
but i feel ppl are missing the point (how can i explain it better?),
all the above suggestions are far inferior visually to what i have proposed (they may perform way better in real life though, speed tradeoffs, gotta love em http://www.opengl.org/discussion_boards/ubb/smile.gif)

but anyways im interested only in a lighting/shadowing method that can achieve the following.

-softshadows from everything attenuated by distance
-bumps (ie everything) casts shadows

(the unreal screenshots dont have either of the above)

the reason is i dont wanna spend my time reasearching something that has been discovered long ago.

cheers zed

henryj
11-19-2002, 06:08 PM
You want us to tell you if what you are doing has been done before, but you wont say what you are doing?

You could submit a paper to a graphics journal. Though this means you have to actually do research which you say you can't be arsed doing so you're screwed either way.

The best plan for you (if you live in USA) is to patent your idea, then sue everyone else who has ever done soft shadows. That way you don't have to do any work and just live off the law suits.

V-man
11-19-2002, 08:02 PM
Let's try to guess.

Find silouette.
Use bumps from bumpmap to create a bumpy silhouette.
Create shadow volume from bumpy silhouette.
Use some shadow blurring trick.

So it's a combination of 3 concepts.

The shot from Unreal seems to have this if you look around the arcs (the one on the right shows well). If they say they have dyncamic lighting all around ....

I think you should create some fancy game and use your idea in it. Better than giving it away! Think like Picaso dude!

V-man

davepermen
11-19-2002, 08:19 PM
hehe, zed, you're now the second paranoid dude i see in here (same forum even?) who thinks the people here are just to steal new algorithms..
if you don't show us what you do we can't help you any further. just remember all the companies that stated yet they found a way to zip lossless muchmuch bether.. they never explained how.. and never had a working app.. http://www.opengl.org/discussion_boards/ubb/biggrin.gif
if you share it, and it's new, it will go into history as zed's bump-shadow method, or something like that. like carmacks reverse.

but if you don't share you don't actually find out if its new or now. we don't list you all sort of possible ways of doing shadows so that you can pick yours out..

the pic looks promising..

Zeno
11-19-2002, 09:06 PM
If people are interested in Unreal 2 shots, check out the 1.5 hour nvidia launch video....Tim Sweeney shows off some footage with soft shadows and dynamic lighting from that engine towards the end.

-- Zeno

pocketmoon
11-19-2002, 11:13 PM
Originally posted by V-man:
Let's try to guess.

OK http://www.opengl.org/discussion_boards/ubb/smile.gif

Of course the danger Zed has placed himself in now is that we come up with a way to do it, maybe even his way, and get the credit forever http://www.opengl.org/discussion_boards/ubb/wink.gif

Well there are two ways to do shadows
a) Shadow Maps
b) Shadow Volumes

Using 'b' would require shadow geometry including the bump so some sort of displacement mapping ?

Using 'a' you could use a shadow map combined with z-correct bump mapping. Capture the scene depth from light POV but replace the depth with fragment depth offset by bump height(hmm another Cg/NV30 demo idea!

That way your shadow map contains z-correct bump info. Then when you do the depth compare, test the current frag depth-bump height with shadow map z. Of course the actual bump height may vanish in a 16 or 24 bit precision bump map but we could use a 32 bit floating point buffer.

Of course this doesn't capture bumps around the edges but at least it would shadow them.

As for soft or anti-aliased or distance-effect shadows - I've already published a way of doing it with depth peeled shadow maps and other folk have show how to do it with vertex weights.

Perhaps none of us is thinking outside the box yet.




[This message has been edited by pocketmoon (edited 11-20-2002).]

pocketmoon
11-19-2002, 11:31 PM
Originally posted by zed:

-softshadows from everything attenuated by distance
-bumps (ie everything) casts shadows


I would guess that the first effect requires shadow volumnes.

I would guess the second requires shadow maps or displacement maps http://www.opengl.org/discussion_boards/ubb/smile.gif

Have you invented shadow volume displacement mapping ?

http://www.opengl.org/discussion_boards/ubb/wink.gif

All I suggest is go for it. Write it up as a paper and stick it on the web. Tell lots of people about it. Nobody is going to steal your idea - although you have to feel sorry for the chap that published 'Carmacks Reverse' a year before JC did in an email!

tayo
11-20-2002, 12:32 AM
I am not convinced by the pictures.
show us a picture with a non planar receiver,
and a more complex occluder.
And i don't see bumps that cast shadow on the picture.

tayo
11-20-2002, 12:37 AM
If you are interested in soft shadows,
you can go to this web page : http://www.mpi-sb.mpg.de/~brabec/

There is a paper on "Single Sample Soft Shadows Using Depth Map"

Zed, if u want to become famous, you will have to write a research paper like this one...

Tom Nuydens
11-20-2002, 12:56 AM
Pocketmoon's ideas sound pretty good. Michael McCool published a technique for reconstructing shadow volumes from shadow maps. Combine that with displacement mapping and you're on your way.

Here's a few papers that I found in my collection and that may or may not be relevant:

"Shadow volume reconstruction from depth maps" by McCool;
"Bump map shadows for OpenGL rendering" by Kautz, Heibert and Daubert;
"Interactive horizon mapping" by Sloan and Cohen.

-- Tom

Nakoruru
11-20-2002, 06:28 AM
What new technique have you presented zed?

I do not understand why everyone is being so unsceptical and undemanding.

I do not understand why zed is being so paranoid.

Zed should just tell us what he has done, no one is going to make any money off of his idea (including himself) whether it is original or not. We should not entertain his fantasy unless he provides a description or a demo.

BTW, I've invented a new techinique that implements the Doom 3 lighting equation on a TNT2 (EDIT: In one pass!). Has anyone done this before? I'm afraid to reveal any details because someone might steal the idea and make a million dollars without giving me credit. Give me a break ^_^


[This message has been edited by Nakoruru (edited 11-20-2002).]

davepermen
11-20-2002, 06:41 AM
pocketmoon. instead of simple zcorrection of the bumps, bether displace them afterwards (should be possible somehow with a postimage process) so that you shift the depthvalues into space..

then again, you could do that on the real screen as well, and you would get displacementmapping http://www.opengl.org/discussion_boards/ubb/biggrin.gif

Matt Halpin
11-20-2002, 06:46 AM
Some more ideas:

Use deferred rendering (ie, render fragment position and normal into 2 render targets and do lighting by a full-screen quad with some funky fragment program) except perturb the fragment position using the bump map. (Is this what pocketmoon suggests?)

Another alternative:
Somehow use the depth replace instruction to change the depth of the shadow volumes per-pixel as you render them (front and back faces). hence the volumes will fail ZTest at different points around the shadow volume. Map the silouette texture uvs along the entire volume so the correct part of the bump map appears aroung the shadow volume...

tempyjoe
11-20-2002, 09:06 AM
Zed i'm not trying to thread hijak, however Epic have said that those shadows are exactly what you have described, YES the bump maps DO cast shadows. I suggest you watch the video.
http://event.on24.com/eventRegistration/...=&format=&key=8 (http://event.on24.com/eventRegistration/EventLobbyServlet?playerheight=533&playerwidth=785&sourcepage=null&eventid=1179&sessionid=1&username=&partnerref=&contenttype=&format=&key=8) 818A278CD4B677B742470856670B9CB

zed
11-20-2002, 09:10 AM
>>Of course the danger Zed has placed himself in now is that we come up with a way to do it, maybe even his way, and get the credit forever

Well there are two ways to do shadows
a) Shadow Maps
b) Shadow Volumes
<<

>>and get the credit foreve<<
http://www.opengl.org/discussion_boards/ubb/smile.gif not very likely cause im using C/ none of the above

>>I am not convinced by the pictures.
show us a picture with a non planar receiver,<<
the sphere with a thing out the top is also receiving shadow (look closely its self shadowing)

cheers for the links tom, IIRC ive seen one before ill cherck the other 2 out

>>no one is going to make any money off of his idea (including himself) whether it is original or not<<

sure it is a bit of vanity but i am worryed about the 'marching cubes' syndrome, money is the least of my worries (if it was i would be in a job that pays better than 5$ US an hour http://www.opengl.org/discussion_boards/ubb/smile.gif ),
btw got any screenshots of the doom3 lighting done on your tnt2? http://www.opengl.org/discussion_boards/ubb/wink.gif

here are some screenshots i done yesterday http://uk.geocities.com/sloppyturds/kea/kea.html
i didnt want to publish these cause they show errors (when u zoom in) that may aid in discovery of this technique,
btw ive also got the technique working on quake2 models (but worse visual results which i do know how to correct but will take quite a bit of time)

ive decided to stop work on this for the time being (KEA is my main priority i should really be finished first before i devote my enery elsewhere)
seeya guys have fun.

ps say i did explain this technique here, whats to stop some guy filing a patent or something saying they invented it?
sure i could say (in court) yeah but look at the post i made in www.opengl.org. (http://www.opengl.org.) (and get laughed out of court i assume).
basically all i want to is
'do it the right way'
if anyone has information on what this is can u email me at sexybastard@xtra.co.nz
cheers

davepermen
11-20-2002, 09:49 AM
whats the power of having a ****ing patent? you want be as useful as the forbitten xor mousecursor years ago? thats utterly stupid, dude. how would you feel if blinn or phong would have been patented, and all other methods. its just another rendering method you present, and you're arrogant and ignorant trying to patent an algorithm. sorry dude, but thats simply stupid. you can get credit, like others do. but ****ing share your knowledge instead of trying to take others away a possibly good way of solving a problem. next carmack comes and patents shadowvolumes, then one comes and patents hw shadowmaps. and then? come on..

can i patent rastericing? i mean.. the algo, not a specific hw implementation..

oh, and raytracing as well..

sorry zed. but a) you should not post in here "has this been done before?" by not explaining what you did. (just from the pic, yes, that has been done before.. as long as you don't give at least hints how you do it, no way of knowing if that has been done. you get it?)
and b) if you don't want to show us how you do something, then why are you in a discussion board?

to all others: find out how he has done it, we can hack him like divx did to mpeg4..

hehe http://www.opengl.org/discussion_boards/ubb/biggrin.gif

seriously, zed, bether make some movies, demos, and presentations about it, with your name everything so we would never forget you...

what if phytagoras would have patented his formula.. how would you normalize normals?

davepermen
11-20-2002, 09:50 AM
btw, i can't watch your page due too much accesses.. please upload your pics on http://davepermen.net on /free, there we have enough bandwith to link here in all pictures..

thanks..

zed
11-20-2002, 10:26 AM
>>and you're arrogant and ignorant trying to patent an algorithm<<

if youll read my post you will see this is NOT what i want to do (but what i was afraid of happening from OTHER ppl)
am i not making myself clear enuf, sorry i do have a difficult trouble explaining myself in english (or any other language, its a fault ive always had,though musically i dont suffer from any problems)

>>, YES the bump maps DO cast shadows. I suggest you watch the video.<<

thanks for the info tempyjoe this does sound more powerful than what i originally thought the epic method was (ie i thought it was similar to the doom3 one),
unfortunatly i have a very slow dialup (countrbumpkin) + cant download video's but i will take your word on it.

davepermen
11-20-2002, 10:33 AM
hm.. how much stolen patents do you know here in opengl.org?
at least give some info how you do it. the idea.. (and dont forget to upload the pics chez me..)

pocketmoon
11-20-2002, 10:37 AM
Zed -

You're going to get nothing but a hard time here unless you spill the beans http://www.opengl.org/discussion_boards/ubb/smile.gif

Are your bump generated via the usual texture bump map or is it a high res mesh ?

V-man
11-20-2002, 10:45 AM
Zed,

will you please give me the algorithm so that I can implement it.

Damn it Zed, why dont you just post the damn thing and get it over with.
What are you afraid of?
Just take the code, press Ctrl+A, then Ctrl+C, then come here and do a Ctrl+V
and feel better.
It's just that simple.
No, Im not gone steal your idea and implement it myself. I just want to see it.


V-man

V-man
11-20-2002, 10:47 AM
Zed?

post it here, OK? Im gone come back in 1 hour.

V-man

Ysaneya
11-20-2002, 11:09 AM
Zed, you still haven't answered my questions :) You make it sound like it's the ultimate method to do soft shadows, when there's maybe pretty bad limitations you don't even see..

I know how you feel, because _fast_ soft shadows is more or less the holy grail of any realtime graphics programmer; but i think others have good points. Can you even patent an algorithm ? The implementation of it, sure, but the idea ? And do you really want to hire a lawyer to do it ? Is it really worth it, knowing that anyway you won't make any $ from it ? Is it really important to become famous (if that's your motivation?) ? Won't the satisfaction to see all graphics programmers, pros or no, use your method be enough (if this is really the "big" thing like you make it sound like) ?

Y.

Mezz
11-20-2002, 03:14 PM
Tell you what zed, print out some code, and then mail it to yourself. The post office date stamp holds up in court.

Then if you want to, you've got proof it was yours to begin with.
If you want to patent it go ahead, I don't have a problem with a patent as long as you say something like "I don't mind people using this stuff, I won't charge them for it".

-Mezz

fresh
11-20-2002, 04:20 PM
That's an old myth. It doesn't hold up.


Originally posted by Mezz:
Tell you what zed, print out some code, and then mail it to yourself. The post office date stamp holds up in court.

-Mezz

FoZi
11-20-2002, 05:16 PM
I think its just a question of time until someone finds out an elegant and fast way to do the same thing, by exposing your technique you will get a lot of feedback in ways that can make your technique even better, that's the whole point i guess. If you dont, maybe someone else will, and that person will benefit from that exposure.

Maybe some part of it is already documented somewhere, or at least a part of it, i remember seeing a very interesting document:
\"Precomputed Radiance Transfer for Real-Time Rendering in Dynamic, Low-Frequency Lighting Environments\" (http://research.microsoft.com/~ppsloan/)
by Peter-Pike Sloan, Jan Kautz, and John Snyder, To appear in the Proceedings of SIGGRAPH2002, July, 2002

...where objects cast soft shadows on themselves and on others, very interesting read, the videos are amazing also. Maybe someone finds it usefull.
About the initial topic, its common knowledge that are many ways to do the same thing, therefore we cannot tell you if your technique has been documented just by looking at it.

Remember, the knowledge belongs to the world http://www.opengl.org/discussion_boards/ubb/wink.gif

[This message has been edited by FoZi (edited 11-20-2002).]

zed
11-20-2002, 06:05 PM
i think ive gone insane.

i keep on reading that ppl get the impression that i want to patent this but i have not recollection of writing this, in fact i swear ive being writing the opposite (can someone point me to the text where i wrote i want to patent this/make money off it etc),

for the final time in simple english (if u dont understand this please email me)
senerio - say it is a brilliant idea, + i post it here.
+ someone else steals it, patents it makes it propriety whatever, this is bad right (this does happen in the real world btw).
now to stop this happening i want to do the right thing
(what it is ive got no idea thats why im asking, do i write up a paper + to a certain specification and give it to some standards comittee + they release it, thus the idea is out in the open? (+ thus cant be made propriety by someone else, ie they very much cant patent something thats already been proved to be invented)

Ysaneya im not 100% certain of everything as im still working on the method (actually this was my original question, has a method like this been done before? the reason i asked this was if it had i didnt wanna spend time on it)

>>Soft shadows do not come without a cost. Is your method fast ?<<

depends on a low polygon scene i can see stencil shadows being quicker but in higher polygon scenes the roles migth reverse

>>What kind of hardware does it require ( GF3/GF4 only ? )<<

im trying to do it on a gf2 at the moment (but cant say for certain what hardware will be required

>>What about the shadows quality ?<<

it is adjustable

>>Is there some artifacts ?<<

yes, though these will hopefully disappear with better quality hardware.

>> From your screenshot it looks like you're not using the distance to the occluder to sharpen it ? <<

yes distance does come into it see the top shadow is sharper than the shadow on the ground. note in this + the other shots the 'shadowspread' with distance mightnt be 100% natural, but this is a minor point that can be easily altered

rgpc
11-20-2002, 06:35 PM
Originally posted by fresh:
That's an old myth. It doesn't hold up.



Does in this part of the world...

---------------------------------------------

Zed, It's a bit difficult for anyone to answer your original question of "Has it been done before" as you are only telling/showing us the end result, not the method.

Sure, I've created self shadowing, bump mapped, textured objects with bone animation and with smooth shadows. Heck I've even done it with 1.2 standard Opengl - probably even 1.1 if I look carefully!

Great, you say, lets see it. I dig out an anim from my lightwave directory and show it to you using my GL avi code.

If your worried about someone "claiming" your idea, as JC apparently has done regarding Carmacks reverse, think about how JC has benefitted from it - he's written a game that uses it - the other guy probably isn't all that famous because all he did was talk about it.

My 2c.

Ysaneya
11-20-2002, 11:14 PM
Hum, if you can make soft bumped shadows run a GF2 faster than shadow volumes on complex scenes, i'd say you are truely on something big :) Although i'll wait to see a demo or some code to believe it.. :)

Does it handle point lights too ?

Y.

pocketmoon
11-20-2002, 11:23 PM
Originally posted by zed:
i think ive gone insane.

You're certainly driving the rest of us mad http://www.opengl.org/discussion_boards/ubb/smile.gif


To answer your question - The BEST thing you can do is publish your method. You can do that here, by email, by submission to Siggraph (although you you will have to produce proper 'academic' paper). The important thing is to put the algo into the Public Domain. Publishing the idea on your own web site should be fine!

No-one can inforce a patent based on something that's already public knowledge (Prior Art).I've been through the patent process myself (in the UK) and basically if someone wants to patent something they can. There is nothing stopping you applying and being granted a patent on shoes and demanding a license from every show maker on the planet. As long as your application is technically correct it will be granted. The enforcability of the patent is another thing altogether and you shoe patent claim would be throw out by a judge.

IMO Your are being paranoid in the extreme and if you're that worried about IP you should take this thread to a legal forum.

What's p1ssing off most folks here is you say 'Has this been done before?' without actually telling us what *this* is! Going on about 'shadows this shadows that' is meaningless! Give us the details or go find a lawyer.

pocketmoon
11-20-2002, 11:44 PM
If your worried about someone "claiming" your idea, as JC apparently has done regarding Carmacks reverse, think about how JC has benefitted from it - he's written a game that uses it - the other guy probably isn't all that famous because all he did was talk about it.


An interesting issue. In the 'old days' advances in rendering techniques would have come exlusively from the academic community. Now we have guys making games with global internet community surrounding them. It used to be the academics who gained recognision through the publishing of papers in journals. Now JC emails MK in May 2001 saying 'hey! Look what I've found' and it's public domain within months. Sim Dietrich presented the zfail approach the year before but he's unknown outside academia.

To be fair to JC, he never asked for the zfail approach to be called Carmacks reverse, but anything which brings qudos to the games industry is a good thing.

I'm hoping that my depth-peeled shadow volumes become known as James Shadows but I seriously doubt it http://www.opengl.org/discussion_boards/ubb/wink.gif

tarantula
11-20-2002, 11:57 PM
Over 50 posts and nothing really interesting http://www.opengl.org/discussion_boards/ubb/frown.gif.
zed, no one's gonna steal your idea. If you still choose to be paranoid you can tell your idea to someone whom you trust on this board to verify so that it will answer your question : "has this been done before?"

GeLeTo
11-21-2002, 02:20 AM
Erm, AFAIK if you make something public domain it cannot be patented (except for the person that disclosed the idea who has 1 year to patent it). Or rather it might be patented but the patent can't be enforced. Making something public domain includes posting your idea on a PUBLIC forum. Or publishing it on a PUBLIC web site (like flipcode).

If you are still not convinced - you can just file a provisional patent application (~75$). Read mere about PPA here: http://www.inventionconvention.com/ncio/specialreport/007.html

In short if you file a PPA and publicaly disclose your idea - it becomes unpatentable by everyone else and you have 1 year to patent it (same as if you didn't file a PPA).
If you file a PPA and keep your idea secret (or you can't prove that you publicaly disclosed the idea) - you can still claim your patent say 5 years later (even if someone filed a patent after your PPA), but the filling date will be established by the PPA.

IMO filing a PPA is paranoid - just post the damn thing here.

[This message has been edited by GeLeTo (edited 11-21-2002).]

LaBasX2
11-21-2002, 02:38 AM
I must admit that I can understand zed's fears. If he has really found out something revolutionary and big he wants of course that he is considered as the inventor of that idea and that is absolutely ok in my opinion.

Of course I would, like all the others here, also love to know how you are doing it. And if I have understood you right, you want to share your idea but you want to make sure that you are known as the one who invented it and thus protect it from being patented by others.
So in my opinion, it is probably the best to write a paper/tutorial/document where you describe the technique and put your real name onto it. You can release that paper on your web page and everybody who will read it will see that you have done it and so you will be known as the inventor of your idea.

But I wouldn't wait too long because else it is well possible that someone else will do something similar or even better and then your invention is probably useless....
I think that you also don't have to make your technique perfect before you can release it. If you have a basic idea, that's already enough to become (more or less) "famous". The original shadow volume approach for example also wasn't perfect at all and still its inventor is named in many books and papers.

LaBasX2

Julien Cayzac
11-21-2002, 02:47 AM
Originally posted by zed:
+ someone else steals it, patents it makes it propriety whatever, this is bad right (this does happen in the real world btw).
now to stop this happening i want to do the right thing
(what it is ive got no idea thats why im asking, do i write up a paper + to a certain specification and give it to some standards comittee + they release it, thus the idea is out in the open? (+ thus cant be made propriety by someone else, ie they very much cant patent something thats already been proved to be invented)


If you posted your method in this forum, we could all attest you were the original inventor (+ if you posted it on the web it would be referenced by google and other crawler... what's the name of that site on which we can browse the web as it was at a given date? I've forgotten http://www.opengl.org/discussion_boards/ubb/frown.gif )

Close sourcing a project / an idea is the best way for it to get stolen in the end with no credit, IMHO...

Julien.

SirKnight
11-21-2002, 04:00 AM
Um....sure this...has been done before. Now post your code. http://www.opengl.org/discussion_boards/ubb/biggrin.gif

-SirKnight

PH
11-21-2002, 04:30 AM
If you guys really want to figure it out, then maybe you should look into two things: displacement mapping and the approach with bumps casting shadows ( paper mentioned somewhere in the thread ). Doing a bit of research won't hurt http://www.opengl.org/discussion_boards/ubb/smile.gif.

knackered
11-21-2002, 05:12 AM
How on earth can a bump map cast a shadow? (I'm not talking about the selfshadow hack)
It is just a collection of direction vectors, so say you have a pyramid bump map, how could you calculate a pyramid shapes shadow from that?
This is all beginning to smell a bit fishy...

Gorg
11-21-2002, 05:19 AM
Lots of people have been doing Soft Shadows(with shadowing of bump maps), so I do not believe you found the holy grail, probably just a derivation of a known technique.

You have a few options : you can dig all existing shadow papers and look if your technique is similar http://www.opengl.org/discussion_boards/ubb/wink.gif

Or you can post it here and see if people can find the root of your technique.

As for getting you idea stolen, this is bogus today. This is the internet. You put something here in matter of days the news will be around.

There are lots of people on this board that have connections, so you can be sure that the knowledge will go around quickly(if your idea is interesting).


[This message has been edited by Gorg (edited 11-21-2002).]

PH
11-21-2002, 05:29 AM
Originally posted by knackered:
How on earth can a bump map cast a shadow? (I'm not talking about the selfshadow hack)
It is just a collection of direction vectors, so say you have a pyramid bump map, how could you calculate a pyramid shapes shadow from that?
This is all beginning to smell a bit fishy...

This is quite old really. Bumps are capable of shadowing other bumps. The technique is a couple of years old ( I even made a post about it - trying to convince Cass to write a demo http://www.opengl.org/discussion_boards/ubb/smile.gif ). Combine that with displacement mapping and it might look 3D.

Edit: Here's the link to the paper,
http://www.mpi-sb.mpg.de/~jnkautz/projects/shadowbumpmaps/ShadowBumpmaps.pdf

If zed shows a screenshot of the knight model casting soft shadows, then I'll be very surprised. Anyway, nice work zed.

[This message has been edited by PH (edited 11-21-2002).]

davepermen
11-21-2002, 05:50 AM
has this been done before?
look, that have i made: 42
has this been done before, i ask you?

oh, i can't tell you with what math formula i calculated 42, it could be the holy grail, you know.. so i ask kindly, has this calc been done before?


you know now how ridiculous your post is zed? not only that all are very interested on how you have done it, you _ASK_ us if we know if your method has been done yet..


similar question..

void RegisterCombinerSetup() {
//code is in here, i took out it's too long..
}

whats wrong with my setup?

http://www.opengl.org/discussion_boards/ubb/biggrin.gif


i bet it has been done before in some or another way, at least rendering packages can render soft shadows, with bumpmaps casting and throwing shadows.. no problem there.. but if they use a similar way you do? dunno..


there are quite some ways.. one i know with raytracing (but for sure mapable onto hw.. is from some parker.. "Single Sample Soft Shadows", a very nice method.. softshadows for realtime raytracers.. small speed drop, high quality, nearly non visible errors..

another method i've read is about rendering line light sources. something in the way: rendering shadow map for the 2 endpoints of the linelight, do some filtering to determine the edges on both images, and with the differences generate some penumbravolume. render this..

there are quite some funny ways.. present us yours.. in some way.. just at least an idea.. you said not shadow volumes, not shadowmaps.. what then? raycasting over the screen image? http://www.opengl.org/discussion_boards/ubb/biggrin.gif

SirKnight
11-21-2002, 06:03 AM
Wow that paper is awesome! http://www.opengl.org/discussion_boards/ubb/biggrin.gif I never knew you could make bumps on a bumpmap cast shadows. w00t! http://www.opengl.org/discussion_boards/ubb/wink.gif

-SirKnight

knackered
11-21-2002, 06:12 AM
Originally posted by PH:
This is quite old really. Bumps are capable of shadowing other bumps. The technique is a couple of years old ( I even made a post about it - trying to convince Cass to write a demo http://www.opengl.org/discussion_boards/ubb/smile.gif ). Combine that with displacement mapping and it might look 3D.

Edit: Here's the link to the paper,
http://www.mpi-sb.mpg.de/~jnkautz/projects/shadowbumpmaps/ShadowBumpmaps.pdf

If zed shows a screenshot of the knight model casting soft shadows, then I'll be very surprised. Anyway, nice work zed.

[This message has been edited by PH (edited 11-21-2002).]

Thanks for the paper link - interesting, but don't see much of an image quality gain vs expense... http://www.opengl.org/discussion_boards/ubb/smile.gif

*Aaron*
11-21-2002, 08:03 AM
Here's my advice: forget about this shadowing technique; it's probably been done before. You should be focusing all your efforts on your game so I can play it. http://www.opengl.org/discussion_boards/ubb/wink.gif

But seriously, if you want to get credit for the technique (assuming it is new) you need to publish in an academic journal. This involves doing a literature search to prove the novelty of the technique. This will require a lot of work and access to a university library. The easier option is to "publish" a simple paper explaining the technique on a CG website or your own website.

Matt Halpin
11-21-2002, 08:04 AM
Heh, I'm using a similar technique to that paper except I'm using a cone to represent per-pixel bump-map visibility rather than an ellipse (It only needs 1 texture, 1 dp3 and 1 add instruction to compute soft shadows).

It's a subtle effect but it makes surfaces look less plasticy...

ToolTech
11-21-2002, 10:39 AM
A simple method I have been using...

1. Use shadow volumes
2. Render shadow volume sides with the same bump texture as the intersection between front cap and sides and clamp them right to bottom cap.
3. Use bump texture to offset stencil z values.

4. render stencil volume as usually and you get the bumped siloutte as a stencil volume side.

zed
11-21-2002, 11:15 AM
>>http://www.mpi-sb.mpg.de/~jnkautz/projects/shadowbumpmaps/ShadowBumpmaps.pdf<<

ive seen this before this method doesnt handle the cases where the bumps cast shadow onto other objects.

>>If zed shows a screenshot of the knight model casting soft shadows, then I'll be very surprised<<

oddly enuf i was working on the knight last night http://www.opengl.org/discussion_boards/ubb/smile.gif

guys no amount of cajoling or threats is gonna extract the method out of me, from the posts ive seen im confident enuf its unique ( im not 100% sure what unreal is doing ),
im not gonna even begin to pick out the obvious flaws with
"tell us your technique and we'll say if its been done before"

the course of action im gonna pursue is get the technique sussed out 100%, get it on paper, stick it in KEA (great idea for publicity) and then lets the wolves loose.
any replys (abuse) please direct to sexybastard@xtra.co.nz
spotya's http://www.opengl.org/discussion_boards/ubb/smile.gif

edit-remember as i said before i havent managed to do this completely in hardware on my gf2 thus its possible it may not be possible (though with better cards i see no reason why not).
shadowvolumes a very likely a better option for ppl wanting to support more hardware quicker.

[This message has been edited by zed (edited 11-21-2002).]

davepermen
11-21-2002, 09:36 PM
well, zed.. i've worked quite a bit with gf2 hw over a long time, and now i'm having a radeon9700 with ARB_fp.. so _IFF_ you want some help for free, you _CAN_ contact me. i do not want to steal your stuff, my stuff until now was always availible for free if anyone asked for it, i share everything and don't need any "made or invented by davepermen" anywhere.. i'm happy enough to see that i helped, and that people use some of my stuff in their projects. they don't need to mentoin me.

so if you're interested in help, you can contact me. we for sure get it working on my radeon, and hopefully we get it onto the gf2mx even as well..

its your choise.. i only offer help.. (and possibly can even answer the initial question to you http://www.opengl.org/discussion_boards/ubb/biggrin.gif)

SergeVrt
11-21-2002, 11:26 PM
Hi zed,

maybe I'm severely wrong but
take a look here
http://www.delphi3d.net/articles/viewarticle.php?article=deferred.htm

it
seems that approach might allow bump shadows

though the approach is somewhat slow on older hardware.

SergeVrt
11-21-2002, 11:40 PM
Originally posted by SergeVrt:
Hi zed,

maybe I'm severely wrong but
take a look here
http://www.delphi3d.net/articles/viewarticle.php?article=deferred.htm

it
seems that approach might allow bump shadows

though the approach is somewhat slow on older hardware.

also some things as ideas are here
http://www.cgshaders.org/shaders/show.php?id=50

dorbie
11-22-2002, 12:48 AM
zed, that screenshot is not bump mapping, it's displacement mapping with shadows. Displacement mapping with shadow casting has been done.

Although something similar has been done before, you may have a completely new approach, even using a known approach but in hardware or partly in hardware is something you could publish.

If you intend to try to publish this I suggest that you DON'T reveal it here. Write something up and try to get it submitted. Do not underestimate the work involved in writing a paper, and don't assume that Siggraph is the only venue to present your ideas in. There are other conferences with a more reasonable and, in my opinion, less corrupt review process.

If you intend to submit a paper for the first time to a conference like Siggraph on your own I think your chances are slim. You need to reference existing works and have a knowledge of the current state of the art. If you can't answer the question "Has this been done before?" yourself and don't have access to the publications to thoroughly research the background then you're in for a struggle.

Perhaps you could team up with someone with the experience required to write the paper and do it with them. Someone at a local university or even via email with someone who is already published in the same field.

If you're going to proceed then you need to get to work researching what's been done before, here's the result of a guick google, but you need to search databases of published papers or at least get busy chasing down reference after reference. Again your local University may help.
http://graphics.stanford.edu/papers/displace/
http://research.microsoft.com/~hoppe/dss.pdf
http://www.matrox.com/mga/products/parhelia512/technology/disp_map.cfm
http://www.mpi-sb.mpg.de/~jnkautz/projects/hwdisplacement/HWDisplacementGI01.pdf
http://www.tml.hut.fi/Opinnot/Tik-111.500/2002/paperit/kimmo_karhu.pdf


[This message has been edited by dorbie (edited 11-22-2002).]

Nakoruru
11-22-2002, 06:49 AM
Can anyone else name a 3D graphics algorithm which is incumbered by patent besides marching cubes? Sure, there are techniques which are patented by the hardware vendors which they license to other hardware vendors, but, all of them are freely usable by the software vendors.

Even if a hardware vendor was so unethical as to scoop an algorithm out of a public forum and patent it, it would not end up preventing any of us from using it. Even the marching cube patent does not seem to prevent anyone from using it.

This dicussion has become ridiculous. Basically zed has said, "I have an incredible secret!"

He should have kept his mouth shut.

He has given us zero evidence that he has what he claims he has. That he should have 72 responses is insane.

Anyone can say "No, its not that, guess again. No, its not that either, guess again." It is insulting.

Until zed proves otherwise, I think he is a liar and a fraud. We should give him no more credit than we would people who claim to have invented perpetual motion machines.




[This message has been edited by Nakoruru (edited 11-22-2002).]

SirKnight
11-22-2002, 07:01 AM
Ya exactly, he gives us a screenshot of some kind of a thing casting a shadow which is soft, then asks, has this been done before with out any explanation of the algo. LOL! I agree that if zed wasn't planning on telling us how he is doing this then he should have kept his mouth shut and suprised us all in the future with a paper of a new algorithm that does soft shadows and is fast.

-SirKnight

dorbie
11-22-2002, 07:55 AM
All the guy is doing is asking a question, give it a rest. If you want to post something more than zed on the subject go ahead write some code.

It's completely understandable that he doesn't want to reveal all of his idea yet and wants feedback on it. This happens all the time in personal communications and discussions inside companies or in academic circles, the problem is this is a public forum where there can be no trust. In the absence of the environment many researchers benefit from he's doing the best with the circle of experts he knows.

SirKnight
11-22-2002, 08:35 AM
Originally posted by dorbie:
All the guy is doing is asking a question, give it a rest. If you want to post something more than zed on the subject go ahead write some code.


Ya obviously. But what kind of question was that? Showing us a pic of a soft shadow and saying if it's been done before? WTH!? I mean come on man. Plus, I never freaking said _I_ wanted to or could design something more than what he did/does. I sure don't remember saying that at all.


-SirKnight


[This message has been edited by SirKnight (edited 11-22-2002).]

SirKnight
11-22-2002, 08:45 AM
Ok you know, I bet zed now through hindsight can see how he worded his first post and realizes it should have been done differently. IMHO I think it would have been better to say, ya I _think_ I discovered a new way of doing softshadows and I don't know if it has been done before so could anyone post some links to papers that talk about blah blah blah? I think saying it like that would not have caused this much of an uproar and after some research he could figure out if what he is doing has already been done. A few hints on things would have helped too without compromising his discovery.


BTW, zed, I didn't mean to talk bad or insult you in anyway from what I have said. I don't want any "beef." http://www.opengl.org/discussion_boards/ubb/wink.gif I just wanted to point out that your question the way it was with no description of any kind was a bit silly. http://www.opengl.org/discussion_boards/ubb/biggrin.gif BUT, I do hope what you have discovered is something alltogether new and the new cool way of doing softshadows. It would be nice to have a softshadow algo that not only looks good and all, but be fast.

-SirKnight

Nakoruru
11-22-2002, 10:43 AM
I'm not worried about getting in a 'uproar' because it the damage has already been done. Only the words of our Lord and Savior John Carmark generate threads this long ^_^

It is just that if you look at exactly what has happened here, you would realize that this could just as easily be an April fools joke as a real discovery.

This is a perfect example of how NOT to announce a new discovery. Be prepared to spill the beans, or STFU.

At this point if better well be new and amazing, or else zed is going to be incredibly embarrassed. ^_^

I hope there are no hard feelings.

davepermen
11-22-2002, 11:34 AM
even if he messed something up and his thing is a) old, or b) crap (or c) yet licenced and not for free http://www.opengl.org/discussion_boards/ubb/biggrin.gif hehe ), it doesn't mather for me. zed is okay, no mather what. this post will get forgotten, but his work not..

did he had done anything yet? dunno http://www.opengl.org/discussion_boards/ubb/biggrin.gif

see, dunno if you're a good or a bad coder actually.. http://www.opengl.org/discussion_boards/ubb/biggrin.gif

anyways, can't wait for some infos.. http://www.opengl.org/discussion_boards/ubb/biggrin.gif you know you can contact me..

knackered
11-22-2002, 02:32 PM
It's obvious zed just got excited, and wanted to tell someone who'd understand the gravity of it (other than his girlfriend who probably said "very nice dear...cup of tea?").

Nakoruru
11-22-2002, 05:04 PM
His girlfriend probably complains about him getting excited too soon as well (JOKE!)

He could have cured cancer for all the good it does us for him to tell us he has and then not say how.

I can understand the need to tell people how you might have done something cool, but without backing it up, it is just hype.

WhatEver
11-22-2002, 07:58 PM
If it's already been done before then you don't have to worry about zeds source code. He'll release the source code as soon as he publishes it. He just wants to make sure he gets the credit he deserves for thinking it up. After that, it's all yours...and it will be free.

Who knows guys...you know Phong? Maybe he's not the guy who came up with that algo first.

About patenting(I know you're not patenting zed). No one would patent an idea like this...it would not be benificial to anyone. A lot of people in here have come up with some amazing stuff...but none of the processes to do that stuff is patented.

zed just wants name recognition...and if he truely has come up with something new, well, then he deserves the credit.

GeLeTo
11-22-2002, 11:50 PM
Originally posted by Nakoruru:
Can anyone else name a 3D graphics algorithm which is incumbered by patent besides marching cubes?...

Pixar for instance has a patent on stochastic sampling. They recently used it to sue a competing company out of business - Exluna which got bought by NVidia.



[This message has been edited by GeLeTo (edited 11-23-2002).]

knackered
11-23-2002, 12:52 AM
Hah, pixar are taking credit for fur rendering, if you watch the documentary on the Monsters Inc. dvd.

pkaler
11-23-2002, 10:15 AM
How about tangent space bumpmapping: http://patft.uspto.gov/netacgi/nph-Parse...,&RS=5,949,424, (http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=/netahtml/search-bool.html&r=24&f=G&l=50&co1=AND&d=ft00&s1=5,949,424,&OS=5,949,424,&RS=5,949,424,)

Elixer
11-23-2002, 10:42 AM
You can just make a web page, with the info explaining roughly how you did it (so you can be on record with 'prior art'), and then slap a (C) on it, and then, if you really wanted to, you can (c) the source, then if it is unique (well, seems these days even if it is not unique, you can then do a register on it, so later on, you can sue ATI, nvidia, intel for patent violation! woohoo!) http://www.opengl.org/discussion_boards/ubb/biggrin.gif

After all the idiots gave whats-his-name the XOR patent, the 1-click buying patent, and the lame list goes on and on ::sigh::

Nakoruru
11-23-2002, 10:37 PM
I wonder how many patents there are like the Tagent-Space-Bump-Mapping one which are not enforced? That one kinda caught me by surprise, because nVidia seems to be the one that pushed this method so much right after they started pushing the GeForce 1 and per-pixel-lighting. Certainly they knew it was patented by SGI.

Stacastical sampling patented? I have seen so many things that use it, it is hard to imagine that Pixar sued for any other reason than the express purpose of putting a competitor out of business.

Of course, with a patent, you have that right (like copyright), you can sue whoever you please whenever you please (this is unlike trademark, which requires you sue everybody). (p.s., I am in the US, your milage may vary in other places).