Search this blog

31 August, 2009

This year's buzzword: Image-Space

It shouldn't be news that the quality of publications done at Siggraph has been going progressively worse, in the last years, and other conferences are becoming more and more important, Eurographics, but also Siggraph Asia (see for example, those really neat publications).

A lot of politics, sponsors, pressure from universities, younger reviewers... Nowadays Siggraph is more an occasion to meet people and see what's going on in the industry, than a showcase of the best graphic research on the planet.

I didn't see anything groundbreaking, and a lot of publications were addressing problems that are not, in my view at least, so crucial. Still I don't think this year Siggraph's was bad, and you'll find plenty of coverage of the event online, so I won't write about that.

I have the impression that the non-realtime rendering, and GI in particular, has seen a slowdown recently, but it may also be that my interest shifted away from those subjects, so I don't have a good picture.

To me what's more interesting, at least now, is realtime graphics, and I'm probably more sensitive to publications in that field. At the main conference unsurprisingly, the most exciting realtime 3d presentation was done by Crytek (see this) but I was really looking forward for the papers of the HPG, one of the Siggraph's side-conferences.

Generally, there is some pretty good stuff there, like the Morphological Antialiasing paper, and many others... But you have to filter out the buzz and I was really bothered by some papers that, in my opinion, simply should not have been there. I don't know really why they bother me, probably it's also because I've seen published some ideas in the past, that I didn't bother to publish thinking they would have been rejected anyway, or maybe it's just that I have too many friends in the research community with good ideas and little luck.

Hardware-accellerated Global Illumination by Image Space Photon Mapping. Wow! Let's read...
And what's that? Well, if you've followed any GPU GI research in the last 1-2 years, it's really easy. They're using a RSM for the first-hit of the lights, they read it back in CPU and use that data for normal photon tracing (claiming that the slowest part is the first-hit, so they care about doing only that in GPU), then they splat the photons using... photon splatting.

It's mostly a tech-demo, it would be cool if they published it as such, maybe with better assets it could be a worthy addition to the other demos NVidia has. Maybe they could have published this applied research in NVidia's GPU Gems. But Siggraph?

Why "image space" anyway? And the worst part, why they don't say "RSM" or "splatting"? They cite those works as "related research", and that's it. They don't use any of those terms, they replaced everything with something else that makes it sound better and new... Photon splatting sounds slow, let's use "Image Space", is way more cool (doesn't matter if there's nothing happening in image space there). RSM are well-known... let's call them... bounce maps (genius)!

Image Space Gathering. Even worse! And it came just after the previous one in the HPG conference! It's something really minor, the only application seems to be blurry reflections, and from the images it doesn't look so nice for that either.

The algorithm? Render your image, and then blur it. But hey, preserve the edges using the Z-buffer, and make your kernel proportional to that too. Wow! Don't blow my mind with such advanced shit man!

They say "image space" and "gathering" and in the abstract,they also use "cross bilateral filtering". The idea is simple and little more than a not-so-neat trick, a curiosity with limited applications. But there's the buzz!

I think it would be easy to write a buzz-meter, check for the frequency of some keywords in the abstracts and build a filtering system that intelligently filters all the noise...


Assen said...

"bilateral filtering" at the end of a graphics paper means "we couldn't really get it fast enough, so we'll just do 1/4 (1/16, ...) the work and pretend it doesn't matter". Okay, it was smart and all the first time, now please everybody stop tacking it onto whatever their real point is; we're all grown ups and can decide what to downsample and bilateral up-res ourselves, thankyouverymuch.

peirz said...

You're right that many papers sound more like smart hacks than like actual research.. but I think it still has value. See for example Asia'09 by Kautz et al; you could say, oh that's last year's Incomplete Shadow Mapping paper where they rasterize at receiver points rather than around the VPLs.. But it's an interesting contribution because even if you say "oh that's just another way of connecting some existing dots", somebody still has to do it and make it work. So the tricks they use to pull it off are interesting, and the fact that it's possible and been done makes us less quick to dismiss 'crazy' ideas in the future.. I guess.

Completely agree about the weird naming though, it's a pain to reverse engineer what they mean. Even RSM is really just a G-buffer.

DEADC0DE said...

peirz: no it's a completely different matter. This work you cite is an extension by the same authors to what they did the year before. The extension is interesting, and their ISM stuff is eve more so, for sure, it's not as new as ISM was when they first published, but that's fine.

The RSM+splatting stuff is not an extension to the author's own work, it's a small trick, with very limited applicability, that could be in ShaderX on in a NVidia demo, nothing more.

And what makes it deprecable is that they never use the words "RSM" or "splatting", they claim it's their cool image space stuff.

So it's not only boring to start with, but it's also a rip-of...

Nothing is new, even photon mapping is just density estimation on radiance, it's a very small, very small trick, but it's elegant, smart, widely applicable and robust. And it made a huge impact in the industry. Still if you see the delta compared to what already existed, is very little, but it's that little detail that made the difference.

Here the delta is probably the same, but the results... And again and again, not citing RSM and calling them "bounce maps" is absolutely censorable.

peirz said...

Oh ok then, I thought for a second you were shooting at anything and anybody that is doing some (incremental) stuff on the GPU :)
I misunderstood, sorry.

Anonymous said...

Just a small comment (as I am just lurking your blog sometime).

I do believe that the peer-reviewer(s?) did not do correctly the job then.

I do agree with you that it is quite unacceptable for a research paper if such things are tolerated.

DEADC0DE said...

I'm not 100% sure, but I think that sponsored papers (i.e. by NVidia etc) do not get reviewed

Robin Green said...

Got to agree on your observations about the quality of recent conference papers. The narrow applicability of many of them is so minor it's pointless to even try to digest them. You may also have a point about the age and experiences of the reviewers.

My personal bugbear is the use of Haar wavelets to do interesting realtime things. Just because you can visually intuit about how they may cross-cancel you can come up with fast, completely landlocked techniques that do not extend to ther basis functions. It's not like other orthonormal basis functions don't cross cancel, itt's just you can't draw pretty diagrams to prove it intuitively.

And there are so many other areas of basic research that are seeing astonishing breakthroughs in fundamental ideas. Hopefully some of these will bubble up to the top soon (in the form of a Siggraph paper for me!, Well, I can dream.)

Kiran S said...

"And what makes it deprecable is that they never use the words "RSM" or "splatting", they claim it's their cool image space stuff."

The current version references both of those works. Not sure if thats new or it's always been there.

I think your opinion of this is a little harsh. They have taken two existing techniques and combined them in a manner that keeps the positives of both while minimizing the pitfalls of each.

DEADC0DE said...

Kiran: You're right, I'm harsh because I hate looking at Siggraph's quality going down the toilet.

And true, in the preprint they do list RSM and Splatting as "related work", but they don't describe their technique as using RSM or Splatting.

And why we have to use the Image-Space buzz anyway?