Your daily teleport
January 10, 2013 11:29 AM   Subscribe

"Every day a PHP script picks a random spot on the land mass of Earth. The nearest photo to that spot is posted here."
posted by Iridic (20 comments total) 27 users marked this as a favorite
 
Is this tumblr very new, or is it intentionally setup so that there isn't a way to go back through the archives?
posted by asnider at 11:36 AM on January 10, 2013


Neat. Generating a random location as lat/lon is easy enough, but how does one scrape up a geo-tagged picture that easily ? (Does google image search allow for lat/lon params ?) Digging around seems to hint that you can do something like that on flickr.

OK looking at the links, it's pulling from http://www.panoramio.com/ , so it might as well be picking random pics from that site, rather than saying it picks a random location and finds the nearest picture thereof.
posted by k5.user at 11:39 AM on January 10, 2013 [1 favorite]


"A few years ago I coded up a little PHP toy that would pull up a random spot on Earth. Now imagine running that script every day and archiving the nearest photo on a Tumblr blog. Except you don't have to imagine it, because Stochastic Planet is already doing it! If you tumbl at all, go check it out — once you've seen a few dozen of these you might have a different sense of what the land mass of the Earth is actually like."
posted by Iridic at 11:42 AM on January 10, 2013


asnider: "Is this tumblr very new, or is it intentionally setup so that there isn't a way to go back through the archives?"

Nope. It started 10 days ago.
posted by schmod at 11:57 AM on January 10, 2013 [1 favorite]


Neat. Generating a random location as lat/lon is easy enough, but how does one scrape up a geo-tagged picture that easily ?

It looks like you found the real answer, but I wondered if they might be getting images from the Degree Confluence Project (previously).
posted by TedW at 12:12 PM on January 10, 2013 [1 favorite]


See also: Confluence.org where you use a GPS to find out where whole-number lat-longs meet and you take photographs in all directions.
posted by Wild_Eep at 12:12 PM on January 10, 2013


So we get a lot of ocean pictures then?
posted by Chocolate Pickle at 12:16 PM on January 10, 2013 [1 favorite]


k5: heavily photographed places would appear more frequently if they did it that way. the topology it'd be drawing from would almost be an inversion of the physical world, with cities and landmarks expanding to fill the map and vast, largely empty swathes of land shrinking to almost nothing.
posted by Ictus at 12:18 PM on January 10, 2013


I'm pretty sure the Flickr API allows for searches via geolocation. So you'd come up with a random coordinate (that isn't in the middle of the ocean), then hit the Flickr API, and then you could even get the photo with the highest "interestingness" from the API as well.

It's actually a cool little thing that I bet many could build in an afternoon.
posted by mathowie at 12:27 PM on January 10, 2013


I'd be interested in knowing whether the site's just using an api that allows for searching by geolocation, or if it actually goes through all the trouble of solving the nearest neighbor problem for a randomly selected set of coordinates.
posted by RonButNotStupid at 12:35 PM on January 10, 2013


So we get a lot of ocean pictures then?

From the looks of what's been posted thus far, the oceans have been excluded. If not, I call shenanigans.
posted by IvoShandor at 12:43 PM on January 10, 2013


So we get a lot of ocean pictures then?

The blog description says "a random spot on the land mass of the Earth", rather than "a random spot on the Earth", so it sounds like it's land-only.

I'd be curious to know how the site is picked. In particular, I'd be curious whether the site is picking points that are evenly distributed over the Earth's land mass, or whether it's just randomly picking two numbers for latitude and longitude. The two methods will yield different results.
posted by Johnny Assay at 1:01 PM on January 10, 2013


I'd be curious to know how the site is picked. In particular, I'd be curious whether the site is picking points that are evenly distributed over the Earth's land mass, or whether it's just randomly picking two numbers for latitude and longitude. The two methods will yield different results.

How so?

Values of latitude and longitude are only stored out to a certain precision, which quantizes them. On some level, I think he has to be selecting from a set of points that are evenly distributed across the surface of the Earth.
posted by RonButNotStupid at 1:16 PM on January 10, 2013


Ooops. Nevermind. I misread your comment. I apologize.
posted by RonButNotStupid at 1:18 PM on January 10, 2013


Every day a PHP script picks a random spot on the land mass of Earth.

Of course, the script is supposed to upper-case a Unicode string. That's PHP for you.
posted by benito.strauss at 3:22 PM on January 10, 2013 [12 favorites]


If you look at the links, it's using panoramio, which is a google-earth-linked database of photos of locations.
posted by anonymisc at 3:28 PM on January 10, 2013


A great way of reminding yourself that a great deal of planet Earth is nowhere in particular.
posted by bicyclefish at 4:53 PM on January 10, 2013 [1 favorite]


If it's a random spot on land, then Antarctica should be getting about ten percent of the images, right?
posted by Chocolate Pickle at 6:32 PM on January 10, 2013


Randomly generating a latitude and a longitude is biased toward the poles: A random point near latitude 0 could be anywhere on the equator while a random point near latitude 90 will always be at the north pole.

The correct algorithm is random distribution on a sphere.

I'm guessing the Google/Panoramio API has some call that lets you get the nearest photo to a lat/long.

The fact that there are no coastal/ocean pictures yet suggests they are ignoring points generated over water.

I did an experiment a couple years ago with randomly generated points on Earth and discovered that there's about a 50/50 chance of being within 1 mile of a road. But if you're not within 1 mile of a road, you're probably at least 20 miles from one.
posted by justkevin at 10:45 PM on January 10, 2013 [3 favorites]


This is stunning.
posted by jb at 11:53 AM on January 11, 2013


« Older loop-a-doop-a, loop-a-doop-a, loop-a-doop-a...   |   Meet Michael Forsberg, Conservation Photographer Newer »


This thread has been archived and is closed to new comments