Talk:941: Depth Perception

Explain xkcd: It's 'cause you're dumb.
Jump to: navigation, search

Somebody needs to try this. Couldn't be that hard. 71.178.11.180 21:27, 22 October 2012 (UTC)

Those must have been some tall goalposts if his point of view is above the clouds! -- mwburden 70.91.188.49 13:16, 10 December 2012 (UTC)

Also, the cameras should be mounted on servos so that when the phone is moved or tilted the cameras can follow, so your viewpoint isn't fixed in one direction. -- mwburden 70.91.188.49 13:18, 10 December 2012 (UTC)

That wouldn't work. The entire football field would have to swivel, or else he'd get some wicked image shearing... 108.28.72.186 01:42, 7 August 2013 (UTC)
actually, it might be possible to correct for that, using bipolar geometry. Essentially, you can derive a 3d model from 2 images from different view points. Here is a (very geeky) demontration of what can be done. Watch the end, where they construct a fly-around video from two images of the opera house in sidney. -- 141.101.104.22 21:10, 19 January 2015 (UTC)
Interesting link, thanks, but I don't think the video was generated from only 2 images, there isn't enough information. If you select "Download the Opera House sequence" you can download the original 43 photographs used. 141.101.99.9 14:05, 19 March 2015 (UTC)
You're right. But of course you wouldn't need a 90-degree flyby for this. 141.101.104.43 16:56, 8 July 2015 (UTC)

An updated solution would be to put the two stabilised cameras on quadracopters which are coded to remain a set distance apart. When you want to look left/right it would take a while for the pair of drones to rotate around their centre point but not too long..... Then you could also get a perspective from the height of a giant (up to 400ft https://www.faa.gov/news/fact_sheets/news_story.cfm?newsId=22615) and with their degree of parallax (from whatever value of height and eye spacing you choose). 108.162.250.225 (talk) (please sign your comments with ~~~~)

This is a very cool project indeed! Some hardcore image stabilizing software would be required too, since you would get nauseous if the two images weren't perfectly aligned at all times. But this setup is the only one I could think of that would enable you to perceive the view from the last frame. Mumiemonstret (talk) 08:44, 17 December 2014 (UTC)
That idea would probably work but at one point, the quadcoptors would have issues flying any higher which would limit the view.

Look at this in stereo mode: http://www.fourmilab.ch/cgi-bin/Solar and cross your eyes so you see three images, then hold your hands up so you only see the one, then... I forget...

I used Google News BEFORE it was clickbait (talk) 12:44, 23 January 2015 (UTC)

I used to do that all the time at one time ... until I got a l...ot of things different to do..

I used Google News BEFORE it was clickbait (talk) 12:44, 23 January 2015 (UTC)

Stereo aerial pairs of clouds do exist see the Google search: https://www.google.co.uk/search?q=stereoscopic+aerial+photos+clouds 141.101.98.206 07:33, 19 August 2015 (UTC)

Or you could ride in an airplane. Or stand on a mountain. 173.245.50.174 19:58, 5 September 2015 (UTC)

That would defeat the point entirely. The distance between the two viewpoints is what provides the increased perspective, not the height of the observer. 141.101.80.84 21:55, 16 November 2015 (UTC)
using a plane isn't a substitute for this, but there's no reason adding the element of flight has to defeat it entirely: you could put one cam on each wing tip & get maybe the coolest effect of all..

Likely the reason the right image is shown on the left and vise versa is that there are two ways to fuse stereo images. Either Walleyed, right-to-rght, or Crosseyed, right-to-left. Doing it the wrong way may result in concave faces and other aberrations. 162.158.255.125 15:20, 30 March 2016 (UTC)

Clouds are fractal, small ones up close look the same as big ones far away. So I don't think this would look as spectacular as imagined. 162.158.39.209 (talk) (please sign your comments with ~~~~)

The thing is that our brains only associate the binocular/3D effect with items that are relatively close, and tend to judge sizes accordingly. If something appears 3D to us, we judge it to be a certain distance away (a key function of binocular vision) and from that we also get a rough estimate of its size. That's why if you see something like a star destroyer in 3D in the movie theater, it looks like something the size of a bus hanging up in the general vicinity of the screen. It doesn't look like something miles long, because big things look flat when they look that size. I believe this way of looking at clouds would give a similar effect. The clouds might look 3D, but they'd also just seem closer and smaller, rather than giving you a real sense of their size. 162.158.78.10 (talk) (please sign your comments with ~~~~)

One way to impress with size is to make sure there's something on the "back wall" to contrast with the foreground. Here's a crosseye cloud picture I made from several pictures out an airplane window. The distance in time between these was one second, according to the EXIF data. Because the foreground cloud is so large, the faraway cloud at the center looks like Randall's "mountains" should. I should note this is a zoom/crop of a much larger pair of pictures. DuplexFields (talk) 01:26, 4 October 2019 (UTC)
I put the above photo on my phone and put it in my google cardboard and it worked super well. Totally amazing.