0941: "Depth Perception"

[Jahoclave]

WHY COULDN'T YOU JUST LET THE CLOUDS BE HAPPY?!

[ragingcomputer]

http://www.hasbro.com/hasbromy3d/en_US/

I just bought one of these yesterday! Today I had planned to duct tape 2 cheap cameras together and take pictures of the cats. Now I might have to look at clouds too!

I'll also leave this link here http://stereo.jpn.org/eng/stphmkr/

[Igor83185]

http://www.newegg.com/Product/Product.aspx?Item=N82E16824859001&Tpk=iz3d (don't have [url])

I did something similar in Fallout 3 at the top of the Washington monument, converging the display on the Whitehouse ruins. Passive 3d FTW.

[thevicente]

Very bad weather during summer in Randall's town.

[Aubri]

He should have the stream available to the public...

In favor! This kind of reminds me of http://vimeo.com/23205323, a time lapse video from the Canaries islands. Watching the clouds slosh against the mountain ridge was the first time my brain really understood, instinctively, that we live at the bottom of an ocean of air. I think this would be the same effect, but for space instead of time.

I'd love to see if someone can make this work. I'm trying to think of complications, but nothing is coming to mind yet. Should work just out of the box. Naturally you still won't be among the clouds, but someone with a pilots license could wire up HD cameras on each wingtip and give us a hell of a show.

This would be... remarkably cool. Especially if it was a really big plane (like a 747 or a C-5) and the video run on time lapse.

Now I really want to get the feeds from two webcams a light-year or so apart so we can get a better perspective on our local stars.
Or since we already know their distances, can someone write an app that does this?

Did someone do that already?

[Dellwood]

That's not exactly true. While you may lack depth perception from parallax, you do have other depth perception. Parallax is not the only evidence our visual system uses to synthesize depth perception. There are other things which cue it, particularly focal length (which does so very strongly), relative motion (which does so strongly), and among others the apparent convergence of lines to infinity, which is the source of some very powerful perceptual illusions involving depth and and relative size perception.

Of course I realise all these things; I was being overdramatic. My point was that no amount of technology (or cues) will allow us to see the world in that way.

If you truly had no depth perception, you wouldn't be able to function in a novel visual environment, yet you can and do.

To be honest, I can't think of any more minor disability, but like any other disability it has many indirect affects on our lives that even most of us affected by it wouldn't realise. Having never had binocular vision, I have no understanding of what the world looks like to someone who does. I can only imagine it looks somewhat similar; people who have lost their binocular vision might find it difficult to adjust, and in the past friends have tried to see how I see by catching or throwing a ball while wearing an eye patch over one eye, and they've had to rely on other cues to perceive depth, with varying degrees of success. However, the thought of being able to perceive the depth of a massive object like a cloud is such a completely foreign concept to me, the only way I can imagine it is to think about scaling up a ball of cotton wool to the size of an office tower, but somehow I think that diminishes the potency of the idea...

[dp2]

Conversely, there's tilt-shift photography, which makes real stuff look like toys.
http://www.smashingmagazine.com/2008/11 ... otography/

[planetjay]

Neil Diamond reference? Joni Mitchell reference? I know this as a Neil Diamond song but.. http://en.wikipedia.org/wiki/Both_Sides_Now_%28song%29

[L33tsaber]

Another member of the Lack Of Binocular Vision Club here. Read the comic; thought "That would be cool... if my eyes converged at a normal angle like everybody else's." Damn esotropia.

[Sockmonkey]

A string of Hubbles along Earth's orbit to do this for stars has been on every astronomer's wishlist for years.

We can still take it pretty far with just what we have available to the people of this board though. If a couple members who live hundreds of miles apart with high-speed net connections and good cameras both sent a feed to the same website...

[Aitamen]

Going to throw out that Mediated Reality opens the doors to this in a big way, heh. This is exactly the type of mind-blowing stuff we could get with the EyeTap and the like.

An amazingly enjoyable concept, and a project I'm sure someone will undertake before too long regardless. I look forward to the fruits of their efforts. "Science Fiction soon devolves into Science Fact by virtue of kids who grew up pondering the impossible as just a little further than it is."

[Cletis]

Surprised that no-one mentioned that Randall switched his left and right on the smartphone screen.

[liljohn118th]

A string of Hubbles along Earth's orbit to do this for stars has been on every astronomer's wishlist for years.

We can still take it pretty far with just what we have available to the people of this board though. If a couple members who live hundreds of miles apart with high-speed net connections and good cameras both sent a feed to the same website...

Actually, truth be told you could do the same thing with just one user and a ground based telescope, with the 6 month-apart shots you would still get the 2AU interocular distance. That being said, as alluded to earlier, stars are REALLY freaking far away, so even 2AU might not do much. The planets will move too much during those 6 months to do anything useful with them, so in the end it might not be as impressive as it sounds at first.

But maybe...

[xX17GHDUDE17Xx]

I don't think an Xkcd has ever depressed me so much. I have no binocular vision.

I feel your pain. My right eye is pretty much useless. I've never experienced a 3-D movie that actually looked 3-D instead of just red.

[lewis]

However, the thought of being able to perceive the depth of a massive object like a cloud is such a completely foreign concept to me

I also have no stereo depth perception, but don't lose heart too much - it only works for fairly close objects under normal cirumstances. Wikipedia says 10 meters, with no citation, so it isn't really a factor in seeing the scale of really large things. That said you might find this book interesting: http://www.amazon.co.uk/dp/0465009131. She writes about the experience of perceiving stereopsis after 30 years without. I spent a bunch of time last year reading about stereo vision and talking to ophthalmologists and sadly what worked for her wouldn't work for me - having differing amounts of myopia in each eye as well as misaligned eyes doesn't seem to be solvable without an arrangement of lenses and/or prisms to correct magnification, focus and alignment at the same time.

Sucks for me because I work in film post-production, and stereo is the New Thing, which has made for some awkward moments in job interviews There's a theory that lack of stereo vision can help other aspects of perception though - see http://www.nejm.org/doi/full/10.1056/NE ... 9163511224.

I tried making little screens for each eye once - Tore open an LCD Casio TV and strung the screen out on ribbon cable and fitted it into a custom-fit headset box. That's when I discovered that no compact lens configuration in the world can make a 1 cm distance feel like it's fifty yards away.

That stuff does exist! I have a pair of video goggles (the FatShark branded ones) which work like that. I had to shove an extra lens in there to correct for my shortsightedness - I just cut a square out of an old pair of glasses. Also think about SLR camera viewfinders, which are constructed so that your eye focuses at a natural distance rather than about an inch away where the ground glass actually lies.

[undeadoranges]

Just attempted this by using a friend's camera phone and my own. I started both of them recording, set one down face up, and ran with the other one to a spot about one hundred feet away to record the clouds. I then held the phones up next to each other and crossed my eyes. It actually turned out pretty well, all things considered. I could definitely tell when one cloud was in front of another.

It wasn't an improvement on just looking at the clouds, but it could be interesting with actual cameras and a greater distance.

[toadpipe]

I've always had excellent depth perception, never had difficulty perceiving scale, so this doesn't seem worth while to me at all. At first thought I was hoping the punchline was going to be mocking the guy that can't tell the clouds are big.

[TheInsomniac]

Try skydiving. You will see them from all sides, and occasionally from the inside

Exactly what I was going to say.

[The Old Wolf]

I am envious of people such as Mr. Munro whose minds work this way. Alas, I must be content to sit in the dirt with a stick and scratch "2 + 2 = 5." Beautiful strip.

[Vnend]

Actually, truth be told you could do the same thing with just one user and a ground based telescope, with the 6 month-apart shots you would still get the 2AU interocular distance. That being said, as alluded to earlier, stars are REALLY freaking far away, so even 2AU might not do much. The planets will move too much during those 6 months to do anything useful with them, so in the end it might not be as impressive as it sounds at first.

But maybe...

Exactly. 14 light-minutes isn't enough separation to resolve depth impressions of objects a minimum of 4 lightYEARS away. We use 76mm separation to resolve parallax info up to maybe 30m, or a little less than 400 times the separation. So if things scale linearly, 14 light-minutes separation would allow for stereoscopic parallax out to 3.9 light-days, quite a bit short of what we would need to resolve even Alpha Centari in stereo.

[cellocgw]

So, going a slightly different way with this...

I could swear I've heard a duet version of "Both Sides Now" with either two female voices or the singer actually singing with herself, but I can't find it anywhere on iTunes. Can anyone help me find it?

Maybe this: http://thevoice.lindaeder.com/voice_sum ... review.htm

Also Hayley Westerna has done this song, on PBS and off, possibly as a duet.

[cream wobbly]

Wow, would this really work? the biggest problem I see is isolating one eye from the other. I just tried holding my iphone in front of my eyes and couldn't find a spot where my eyes didn't see the entire screen.

You may have heard of these things called "3D Glasses". I'm not talking about passive lenses (though they would certainly be cheaper). Shutter lenses are the cheapest way to go to provide a full-colour image to each eye, but the headset type of thing envisioned by our artiste here has been in existence for decades now. Here's a fairly recent example for you to lust after: http://www.techradar.com/news/home-cine ... iew-920230

So yeah, it would really work.

What gets me is why our artiste limited himself to the possible. What's wrong with a pair of gigantic robots, each moving in perfect synchronicity with the other, and with the wearer, to provide the illusion not only of depth, but of power?

There are practical applications (already in use, hinted at in the "snaps back" panel) -- there's a technique in astronomy which records a piece of the sky at six month intervals, yielding a spacing between eyes of 2 AU. This gives a real perspective on the Universe, and the best we're likely to get in a lifetime.

[cream wobbly]

Actually, truth be told you could do the same thing with just one user and a ground based telescope, with the 6 month-apart shots you would still get the 2AU interocular distance. That being said, as alluded to earlier, stars are REALLY freaking far away, so even 2AU might not do much. The planets will move too much during those 6 months to do anything useful with them, so in the end it might not be as impressive as it sounds at first.

But maybe...

Maybe actually?

The planets move, but is that all you're interested in? Sheesh.

If you can reliably plot a difference with 2AU, then you can remap their positions and expand the separation so you'd have to go crosseyed to see Alpha Centauri. Cheating? Absolutely. Accurate? Absolutely not. We're looking to create a perception.

[Harold]

I'm going to attempt something similar, just for kicks, sometime when I have a break from uni.

Just to simplify things; I'm going to use two iPhone 4s and a MacBook. Through the wonders of wifi this should be fairly easy to set up, and with a decent AP I should be able to obtain at least 100m separation between the lenses, probably more. Any slight misalignment can be corrected in software, and the results will be viewed in realtime with the side-by-side stereoscopic method.

If this does work out, I'll make some recordings for the wider community to have a peek at. Wish me luck!

[Saylone]

So, anyone actually tried to do this?

All you need is 2 photo cameras and a friend:


Actually, no.
You also need a larger distance (10 meters here), non-sunset lighting (causes colour mismatch, clouds don't look similar enough in R and GB for eyes to lock on to), better clouds, and a place without a construction site in the way...

Take 2, colour removed for clarity, distance increased to 15 meters:


Still kind of wrong.
I guess i'll wait for better weather and lighting.
And we should stand on different buildings...

Simple idea, devil in the details.

[Harold]

Still kind of wrong.
I guess i'll wait for better weather and lighting.
And we should stand on different buildings...

Simple idea, devil in the details.

You missed the point.

The wild distortion (due to the parallax effect) you showed there is only going to effect photos of objects that are relatively close. For things like clouds, where the parallax effect is extremely slight, the 3D effect should be more apparent.

[Saylone]

You missed the point.

The wild distortion (due to the parallax effect) you showed there is only going to effect photos of objects that are relatively close. For things like clouds, where the parallax effect is extremely slight, the 3D effect should be more apparent.

I thought the point was to get the clouds in perspective with the buildings?
Then you'll need a clear view of the city and a large enough separation to see the clouds in volume and feel the distance to them.
For me the interesting part is getting an idea on how far and high the clouds really are, compared to regular cityscape objects.

[SpringLoaded12]

For those asking whether or not this would work, yes, it would. Using contraptions to spread the view from each eye far apart will increase how far you can see. I've seen a diagram of a device like a telescope, but that spread each eye's view apart horizontally by about 8 feet using mirrors and lenses. It was a very old diagram and description, but supposedly it would allow you to see things a mile away with good detail.

[GenericAnimeBoy]

Actually, truth be told you could do the same thing with just one user and a ground based telescope, with the 6 month-apart shots you would still get the 2AU interocular distance. That being said, as alluded to earlier, stars are REALLY freaking far away, so even 2AU might not do much. The planets will move too much during those 6 months to do anything useful with them, so in the end it might not be as impressive as it sounds at first.

But maybe...

Maybe actually?

The planets move, but is that all you're interested in? Sheesh.

If you can reliably plot a difference with 2AU, then you can remap their positions and expand the separation so you'd have to go crosseyed to see Alpha Centauri. Cheating? Absolutely. Accurate? Absolutely not. We're looking to create a perception.

To view the planets, you could place a telescope at the the Earth-Sol L3 lagrange point (with a communication relay at L4 or 5) and you could take the pictures (roughly) simultaneously. You would have to view the nearer telescope with a delay on the order of 45 minutes. This would of course be extremely expensive, but it is quite doable.

Howver, I'm afraid that even the closest stars exhibit an imperceptible (to the naked eye) parallax--only slightly more than roughly 1 arcsecond, even with the image sources separated by 2AU. See also The wikipedia article on the 'Parsec'. Like wobbly said, you would pretty much need to use a computer to enhance the parallax separation.

[nealh]

1. Put cameras miles apart on the Sandia Mountains.
2. Record.
3. Display on UNM's dome.
4. ???
5. Profit

Albuquerque, NM we could make this happen!

[Harold]

I thought the point was to get the clouds in perspective with the buildings?
Then you'll need a clear view of the city and a large enough separation to see the clouds in volume and feel the distance to them.
For me the interesting part is getting an idea on how far and high the clouds really are, compared to regular cityscape objects.

From what I understand of stereoscopy, you'll have a hard time getting the eye to believe such an image. The only way that it will work is by showing the scale of the clouds relative to distant things, or moreover enabling one to see the depth of the cloud. As it stands we see clouds as a flat object, since our eyes are too close to observe that sort of scale.

I could be completely wrong, but I welcome that too

[Saylone]

I've seen a diagram of a device like a telescope, but that spread each eye's view apart horizontally by about 8 feet using mirrors and lenses. It was a very old diagram and description, but supposedly it would allow you to see things a mile away with good detail.

Sounds like what is called a stereopipe here - a combination of binoculares and a periscope for improving depth perception on large distances:


Can't find an english term for that.

From what I understand of stereoscopy, you'll have a hard time getting the eye to believe such an image. The only way that it will work is by showing the scale of the clouds relative to distant things, or moreover enabling one to see the depth of the cloud. As it stands we see clouds as a flat object, since our eyes are too close to observe that sort of scale.

Well, the second of my pictures actually give some sense of depth in the clouds, while still preserving depth on the city, so i guess it's a matter of getting a proper perspective.

[Reg Shoe]

It's a stereoscopic rangefinder.

Onces 10-15 feet across were used in World War I on battleships.

For some reason I can't post images yet (new user?) but google image search for "battleship rangefinder" will show you some pictures.

[TomN]

I'm detecting an Android app that can be run on two phones to gather the video data.

[wingsofwrath]

For those asking whether or not this would work, yes, it would. Using contraptions to spread the view from each eye far apart will increase how far you can see. I've seen a diagram of a device like a telescope, but that spread each eye's view apart horizontally by about 8 feet using mirrors and lenses. It was a very old diagram and description, but supposedly it would allow you to see things a mile away with good detail.

That sounds like an old-fashioned artillery coincidence rangefinder (http://en.wikipedia.org/wiki/Coincidence_rangefinder), which relies on an observer achieving perfect superposition of images from two telescopes with their apertures placed some distance away (usually 1 meter, but can be as much as 3) and then reading the resulting range (calculated by the angle of rotation of the prisms in the eyepieces relative to the distance between them) on the scale of the instrument.

[teryret]

Surprised that no-one mentioned that Randall switched his left and right on the smartphone screen.

I noticed that too. In fact I read this thread just to see if it was explained. My best guess is that the images are flipped for the combination of two reasons:

1) going walleyed is so much harder than going crosseyed (most people can't do it at all, although I personally suspect that with enough training most people could learn to) and
2) in order to put the centers of the pictures within the range of parallel-eyed to slightly crosseyed (the natural range) there would need to be a huge degree of overlap and you'd need some mechanism to avoid getting the wrong parts of the overlap into the wrong eyes (ie the glasses that come with 3DTVs).


I remember back in the day there was a Quake 3 patch that added crosseye stereoscopic rendering so that you could see the 3D game in 3D. It was pretty awesome, rockets coming at you justified the learning curve by itself. It also revealed just how many old-wives tales everyone seems to believe (won't you get stuck like that? doesn't that give you a headache? that's not good for you! etc)

[Black Daemon]

You know, I wonder if you could make two periscopes--one for each eye--then go around using them to create the perspective you want on things around the house! The great thing about this is that the perspective is easily variable: just rotate your periscopes! Could be like ||, or \/, or -- --! The possibilities are neat!

[eidako]

* Stellarium is a planetarium simulation program
* Stellarium is open source
* download Stellarium source code
* isolate the function which renders stars
* replace it with a function which renders stars as an anaglyph, separating the two channels by a distance determined by multiplying a constant (perhaps half the viewing angle) by the ratio of Alpha Centauri's distance and the star's distance
* compile
* upload it for everyone else

Too lazy to do it, with any luck someone else isn't.

[Nexxo]

It's geeky and beautifully poetic at the same time.

[liljohn118th]

Actually, truth be told you could do the same thing with just one user and a ground based telescope, with the 6 month-apart shots you would still get the 2AU interocular distance. That being said, as alluded to earlier, stars are REALLY freaking far away, so even 2AU might not do much. The planets will move too much during those 6 months to do anything useful with them, so in the end it might not be as impressive as it sounds at first.

But maybe...

Maybe actually?

The planets move, but is that all you're interested in? Sheesh.

If you can reliably plot a difference with 2AU, then you can remap their positions and expand the separation so you'd have to go crosseyed to see Alpha Centauri. Cheating? Absolutely. Accurate? Absolutely not. We're looking to create a perception.

I have no problem with cheating! But, the planets also tend to rotate, meaning pictures taken 6 months apart aren't going to get the same relative view of the surface. Depending on the planet and the useful magnification, perhaps no big deal. But you could end up, for example, with two pictures taken of the same side of a planet from different perspectives, which would make the whole 3d thing hard to pull off.

However, I'm thinking two users on opposite sides of the globe snapping at the moon at the same-ish time could produce some useful results.

Experimentation warranted.