Join 3,562 readers in helping fund MetaFilter (Hide)


fake or photo?
April 25, 2005 5:03 AM   Subscribe

Fake or Photo? Can you tell the difference between a photo and a CG render?
posted by crunchland (43 comments total)

 
hmm. too bad i cant see any images
posted by obeygiant at 5:06 AM on April 25, 2005


Apparenty I can tell the difference between a photo and a CG render.
posted by cloeburner at 5:09 AM on April 25, 2005


They wish they were at the point where you can't tell the difference, but no, not there yet. CG images are just too perfect. [14/14 baby]
posted by Plinko at 5:15 AM on April 25, 2005


I thought this was cool. But I only got 8/10.
posted by grouse at 5:16 AM on April 25, 2005


the shot glass in the bonus round was the only one I got wrong.
posted by crunchland at 5:21 AM on April 25, 2005


CG skin still isn't realistic. Especially when it looks like a CG harelip.

I found myself looking for imperfections... If I didn't see any (or didn't see enough) I suspected fake. 8 for 10. The lighting effects in one image threw me. (Must have been some polarized filter or something...)
posted by caution live frogs at 5:22 AM on April 25, 2005


yeah, the light rays bouncing off the table in that one looked too straight...guess they were real. And the chess pieces in the other all seemed too uniformly out of focus...but also real. The one that fooled me was the blue glasses--damn realistic.
posted by zardoz at 5:36 AM on April 25, 2005


What gave it away for the chess pieces was the dust.
posted by Plinko at 5:49 AM on April 25, 2005


I got 8/10. What makes it tricky is that some of the "real" photos look like they've been Photoshopped...
posted by Jon-o at 5:55 AM on April 25, 2005


They still don't have it down. What does that tell you, Hollywood?
posted by password at 5:58 AM on April 25, 2005


9/10 - the first grass fooled me altho' I've had to deal with a lot of shots of synthetic turf in my time so that's my excuse ;-)

Having managed studios with lots of photography & CG I should've developed an eye by now...
posted by i_cola at 5:58 AM on April 25, 2005


They wish they were at the point where you can't tell the difference, but no, not there yet. CG images are just too perfect. [14/14 baby]
posted by Plinko at 1:15 PM GMT on April 25 [!]


Yeah, exactly - 14/14 here, too... the big give away is the small imperfections - dust on the chess pieces, the grime and swirl marks on the black chairs (bonus #3), the fact that all the nails (on the "scatter" pic) have absolutely identical heads...
The starbursts on the table was a good one - I umm'd and ahh'd a bit, and then took a guesstimate; there's water droplets on the chair, and beading forming as run-off from the table top...

To be quite honest, though - all it would take is more attention to detail (or an automated "add imperfection" optin...), and it would be all-but impossible to differentiate between real and CG.
posted by Chunder at 6:03 AM on April 25, 2005


Only 7/10, which probably doesn't qualify me to add that the thing about CGI images is that they're often made by people to show that they can (I guessed the glasses because of that: good CGI, boring photograph). I guessed the table (wrong) because I thought it was someone showing off texture mapping.
posted by Grangousier at 6:12 AM on April 25, 2005


The photos aren't real enough anyway...
They were meant to look like renders. Let's see a real street scene, a park, and a full face. We are still some years off, hope we get there though...
posted by lorbus at 6:21 AM on April 25, 2005


Yeah, but the images on that site haven't changed for over a year and a half. That doesn't exactly make it cutting edge CGI.

Oh, and the scattered screws stopped me getting 10/10.
posted by twine42 at 6:37 AM on April 25, 2005


Double post from 2001, IIRC
posted by Zurishaddai at 6:37 AM on April 25, 2005


These images are almost four years old, so I would't say they're state of the art. And that skin rendering is very flat and dead by todays standards.

That said I still missed three of them. They got me with the wineglass cliche, bendy nails, and fake grass.
posted by crumbly at 6:47 AM on April 25, 2005


14/14.
@ those saying the technology isn't there yet:
when discussing images of these sizes/resolutions, what we have now is totally sufficient. those pictures are small/low rez enough that there are probably artists that could paint or draw something indistinguishable from a photograph if appropriately reduced.
posted by juv3nal at 6:50 AM on April 25, 2005


Yee gads, I'm slow. Anyway check out some of the images done with MaxwellRender. Next Limit is attempting to build a no (ok, minimal) compromise physical modeling renderer. Although it's still in beta I think there's a wonderful quality of light to the renders it produces.

And as artists (and programmers) get the hang of technologies like subsurface scattering, the photorealistic humans will follow.
posted by crumbly at 7:17 AM on April 25, 2005


And there's more where that came from. Although personally I prefer a slightly less photorealistic approach.
posted by crumbly at 7:33 AM on April 25, 2005


crumbly: Pu Yi was scary.
posted by Termite at 7:39 AM on April 25, 2005


It seems I'm not too good at telling the difference between THUMBNAILS...
posted by odinsdream at 7:41 AM on April 25, 2005


I got the first 10 all correct. Missed 1 in the second group, I thought the dice looked fake. To be fair, I do photography and have taken a number of shots like some of these. I also have done CG, and looked at a lot of CG work.
posted by Goofyy at 8:06 AM on April 25, 2005


odinsdream, you should click on the thumbnails to get the bigger version.
posted by donut at 8:07 AM on April 25, 2005


Missed the dice. If I ever do computer rendering, I know now to go in with a fine pencil tool and add dust and scratches.
posted by klangklangston at 8:40 AM on April 25, 2005


10 of 10. Woo.

I dunno: there are all these little details that I just don't see people getting right in CG. Scratches on the chess pieces, the bubbles in the wine, etc. are total giveaways.

For example, the Xbox 360 pic looks like total CG, but CG based on someone who's either seen the real thing or is making educated guesses based on the "leaked" photos.
posted by sachinag at 9:38 AM on April 25, 2005


CGI can always be made to appear utterly indistinguishable from reality, given enough time. Time is the only real limit. Rendering time (for sub-surface scattering, radiosity, HDRI or whatever else), time to design musculatures, time to sprinkle motes of dust all around. Attempts at photorealism that fail are almost always due to taking shortcuts. There's nothing an artist who uses software can't do, given time.

(6/10 the first time on the first round, then 4/4 the first time on the bonus round.)
posted by jiawen at 10:51 AM on April 25, 2005


Is it just me or are the images is in the "bonus round" just loaded with JPG artifacts?

Look at the "chairs" picture in the bonus round, the ground is so blurry and undefined you can't tell if it's real or a texture.

Some of the photos seem made to look more like CG then is natural. The blades of grass in the earth-ball picture look like paint strokes.
posted by delmoi at 12:27 PM on April 25, 2005


reading the comments, I see that this is from four years ago. Makes sense, and explans why the JPGs look so awful.
posted by delmoi at 12:33 PM on April 25, 2005


This is actually a harder contest to do if you know something about cg. I suspect that's who this is really aimed at.

For instance, the golfball appears to be a cg image, since the blur of the grass behind the ball looks like a photoshop filter.

There are other giveaways, too. Here are some things to look for that never failed me in this contest (although I did not score perfectly.) if you want to know where cg and reality still differ, by now:

1. Shadows. How do the shadows react? In almost all 3d programs available today, shadows are expertly done, (although quite expensive for render time when the applications full abilities are used) except for one aspect: Shadow attenuation. Shadows do attenuate in 3d apps, which means they fade away the farther you get from the shadow's source, but not realistically. If you look at the hammer shadow and bent nail image, you'll see that the shadows of the nails widen and blur at the edges MORE the farther you are from the nail, but remain tight and crisp at the base of the nail. Skilled 3d compositors and renderers have workarounds to simulate this effect, but it's not a function of the default maya renderer. If that image were recreated in Maya, chances are that the shadows would either remain crisply defined and be shaped precisely the same way as the nail was, OR they would just fade away, as though the shadow were being gradually erased from it's farthest point toward the source. Neither of these is how shadows actually react.

2. Caustic illumination: If you look at the glass image (I got this one wrong, actually.) you'll notice the way the light bends and makes patterns at the base of the glass. Until recently, this was an EXTRAORDINARILY difficult effect to produce in 3d. Usually you created shaped lights or actually textured light patterns directly on the table to do that. Now, thanks to the widespread use of renderers like Mental Ray, you can do this easily at the cost of render time. Either way, this lighting effect is created in real life when the light passing through or bouncing off of a refractive/reflective object bends when encountering that object. The curves in the water glass are clearly the refractive object in this case. This light pattern is called Caustic Illumination. It's a form of indirect or global illumination, which is what you call it when something is lit purely because the light bounced off a nearby object. To see another form of Global Illumination in real life, go in a dark room with a flashlight and wearing brightly colored clothing. Shine the flashlight at your clothes and hold your hand just outside the radius of the light right against your clothes. You'll see your hand will be lit, but the light will be weak and slightly colored to match your clothing.

3. Bokeh: To my knowledge, Bokeh is the only thing cg is completely incapable of rendering. Bokeh is what happens to light when you take a photo that has a light source out of focus in it somewhere. For instance, if you photograph and person standing in front of a light bulb, with the person in focus and the light bulb out, you'll see that all of the out of focus material will blur, but the light bulb will become a peculiar orb of color. If you look at the photo of the chess pieces, the ones in the back have these orbs forming on them because of the light reflecting off their metal surfaces. Bokeh. This effect does not happen in Maya. That's not to say there aren't workarounds (especially in still images where you don't have to worry so much about light accurately reacting to motion), but it's the best signifier, because most people aren't even aware of it when making their images.

Now, none of these are sure bets, but if you're interested in cg, it's important to know about them, anyway. Like I said, I think this contest was geared, in some ways, to fool people who are familiar with cg. The bokeh was a dead giveaway in the chess piece and patio furniture pictures. On the other hand, flawed geometry and dirty surfaces are the kinds of things that any 3d renderer worth his salt would be sure to include in an image, whereas a great photographer can make any object seem pristine seen through his lens. So there's no sure way to know, if the 3d artist is skilled enough (at least, if you're not talking about people. Few can really make a photorealistic person that stands up to intense scrutiny.) and knows the properties of photography and light. But these are good indicators. Hope this was informative.
posted by shmegegge at 5:26 PM on April 25, 2005


and I clearly came altogether too late to the party. That's ok, though. I still love you all.
posted by shmegegge at 2:46 AM on April 27, 2005


Something else has occurred to me that you all may want to consider while you're busy not reading this thread, anymore. Global illumination other than caustic illumination will generally be a good indicator that you're looking at cg. What I'll describe will tell you definitely that something IS cg, if you notice it, but not necessarily that something is REAL because people with talent can fake these things well.

Like I said above, Global (sometimes called Indirect) Illumination (GI) is when something is illuminated by light bouncing off something else. So when you see a shadow, for instance, and it's weak and not very well defined, despite the fact that no sunlight is falling directly on the spot where the shadow is, that's because light is bouncing off of clouds, trees, you and everything else around to hit that spot indirectly. It has a shadow because it is not directly illuminated, but it isn't pitch dark because it is indirectly illuminated. Now, the nature of shadows is an interesting indicator of GI because the default way that 3d applications create shadows is to create a black shape on the shadowed surface (I'm over simplifying) and these shapes will be very hard edged and dark unless you alter them. It's easy to make them less dark. It's less easy, as I've mentioned above, to make them accurately less hard edged. Now, certain types of light will create very very hard edged shadows that resemble the silhouette of their source object from the perspective of the light source. Any strong light source from a great distance will do this. Generally, that means only the sun. But the moon might do it in a large unobstructed area on a clear night. So if you see a bright sunny day making weak, soft edged shadows... chances are it's cg. Conversely, if it's an indoor scene, with a close light and the shadows are crisply defined and dark... definitely cg.
posted by shmegegge at 3:09 AM on April 27, 2005


the point of the above was not to be repetitive, but to call attention to the nature of outdoor lighting and and indoor.
posted by shmegegge at 3:13 AM on April 27, 2005


This all seems awfully confusing.
It was all so much easier to understand when Jeremy Birn explained it in his awesome book. But then, he's got a degree and plenty of illustrations. Go get that book.
posted by shmegegge at 12:36 AM on April 28, 2005


That's enough out of you. I'm the one teaching cg, here. Not Jeremy fucking Birn. Quit undermining my authority, you!
posted by shmegegge at 12:37 AM on April 28, 2005


But Professor! How would one fake these effects of global illumination that you speak of? You know, to save render time?
posted by shmegegge at 12:38 AM on April 28, 2005


I'm sorry, but I don't answer questions unless the student asking them has first raised his or her hand and then been called on by me.
posted by shmegegge at 12:39 AM on April 28, 2005


*raises hand.*
posted by shmegegge at 12:39 AM on April 28, 2005


Yes... I'm sorry, what's your name again?
posted by shmegegge at 12:39 AM on April 28, 2005


shmegegge.
posted by shmegegge at 12:40 AM on April 28, 2005


Yes, shmegegge? What's your question?
posted by shmegegge at 12:40 AM on April 28, 2005


Well, how would one go about faking these global illumination effects that you talked about? You know, to save on render time?
posted by shmegegge at 12:41 AM on April 28, 2005


Ah, well I'm glad you asked. It's a very interesting question, with an equally interesting answer. Basically... oh, but I see we've run out of time. Next class, I'll get to it.
posted by shmegegge at 12:41 AM on April 28, 2005


« Older Vive la Revolution....  |  A comprehensive catalog of air... Newer »


This thread has been archived and is closed to new comments