Join 3,497 readers in helping fund MetaFilter (Hide)


Astronomers need your help
July 11, 2007 4:46 AM   Subscribe

A team of astronomers needs your help. It's not terribly easy to get computers to distinguish between galaxy shapes, but fortunately humans are not only very good at it, but seem to actually enjoy gazing out in to space. So, go to galaxyzoo.org, look at a few pretty pictures from the Sloan Digital Sky Survey , and help classify millions of galaxies and aid research in to how they form and evolve while you're at it.
posted by edd (43 comments total) 14 users marked this as a favorite

 
The site still claims not to have been officially launched, but since the BBC News story has come out, and there's a Facebook group that wants the word spreading, so I think it's ok to start sending this link about. They're also rather flooded with emails by the sound of it, so if you have questions I'll answer the ones I can myself, and I work with a couple of the team members (but not on this) so I'm happy to pass other questions about the project on upwards so they don't get overwhelmed by MeFites!
posted by edd at 4:49 AM on July 11, 2007


This is really awesome. Too bad that first two pictures I got (after a few timeouts) were, as far as I could tell, indistinct blobs. I just clicked on the button marked 'your mom' and moved on.
posted by Kattullus at 4:59 AM on July 11, 2007


MetaFilter: Just click on the button marked 'your mom' and move on.
posted by Poolio at 5:03 AM on July 11, 2007


A victim of its own success already I guess - it looks like it's the SDSS catalogue server that's struggling at the moment.

If it's an 'indistinct blob' it sounds like it's probably an elliptical, but you can always click 'don't know' if you're not sure, and they'll be submitted to people repeatedly to help deal with mistakes and more ambiguous cases.
posted by edd at 5:04 AM on July 11, 2007


Okay, now I'm getting the hang of this. This is a lot of fun! It's like the stupid games I waste my time playing on the internet, except for a higher purpose.

Oh, and Galaxy Zoo would be a kick-ass name for a rock band.
posted by Kattullus at 5:04 AM on July 11, 2007


And now it's dead. Whenever I go into the galaxy analysis section, all I get is a runtime error.
posted by Kattullus at 5:12 AM on July 11, 2007


I will help you with your astronomology if you will then build me my house in the sky.
posted by Astro Zombie at 5:42 AM on July 11, 2007


Kevin Schawinski, an astrophysicist at Oxford University, UK, is one of the team who devised the project.
"I classified about 50,000 galaxies myself in a week," he said. "It was mind-numbing."
No no, that's your normal state.
posted by elpapacito at 5:44 AM on July 11, 2007


This seems like the perfect kind of thing to farm out to Amazon's mechanical turk.
posted by srboisvert at 6:15 AM on July 11, 2007


Cool. If anyone knows of any other projects similar to this, please let me know or post them at Ponderfodder.com, as I'm trying to amass a long list of this sort of thing.
posted by Zinger at 6:39 AM on July 11, 2007


Is it coincidence in the tutorial section that the eliptical galaxies are more orange in colour while the spiral galaxies are nearer white? Might this assist later when they start asking about small blobs?
posted by nthdegx at 6:48 AM on July 11, 2007


No, not a coincidence. Ellipticals are generally older redder galaxies, and spirals contain younger bluer stars.

It's not something you'd want to use to determine morphology though, as while there's certainly that trend there are other things that complicate it.
posted by edd at 6:59 AM on July 11, 2007 [1 favorite]


Cheers.
posted by nthdegx at 7:04 AM on July 11, 2007


I was going to FPP this. A great site.
posted by WPW at 7:17 AM on July 11, 2007


By taking part, you'll not only be contributing to scientific research....

You only have seventy five more to go...
posted by phaedon at 7:40 AM on July 11, 2007


This is awesome! I heart galaxies.
posted by Mister_A at 7:49 AM on July 11, 2007


astronomy == stamp collecting.

Snark aside, are they running the images multiple times to minimise mis-classifications? What good is the data once it's been generated? How did they identify these particular blobs as galaxies in the first place?
posted by Leon at 8:05 AM on July 11, 2007


Yes, the galaxies will go through multiple times. Plus, I imagine that once classified it's very easy for one of the team to scan through by eye and look for misclassified objects without having to do the whole clicky-thing. In the same way I could show you a load of images and you could spot an odd one out without having to manually sort each one. Generally, they do have a good set of controls to make sure the data that comes out of the end of this process is good.

Once they have the data they can examine in more detail trends like the one nthdegx noted, and could look for, say, objects that don't fit that trend and investigate them further. Tied together with the spectra that SDSS also provides there's possibilities for even more detailed studies, generally with the aim of finding out what makes a galaxy spiral or elliptical, and how the two types form, how they change from spiral to elliptical (or maybe the other way I was surprised to read) and when.

There's more unusual things too - for example, the clockwise/anticlockwise thing is useful because of a suggestion in a recent paper that galaxies on one side of the sky might be more clockwise than anticlockwise, and vice versa on the opposite, which would be very interesting if it turned out to be true.

There's a fairly comprehensive page or two on these things at the Galaxy Zoo site but unfortunately it's down at the moment.

To the last point, galaxies are identified by basically being extended objects, which fortunately there are pretty reliable ways to determine by computer - the Sloan survey does this for every object it detects as a matter of course. It's the determination of structure within that that's tricky to do by computer.
posted by edd at 8:52 AM on July 11, 2007 [2 favorites]



If you really want to be snarky about it, all research is stamp collecting, to start with. Particle accelerators? Just high-energy stamp collecting. History? Dead-tree stamp colecting. Etc.. But I don't mean to presume to take that criticism too seriously.

As for identifying blobs as galaxies, that part is fairly simple. Stars are so small (compared to galaxies) that they will all show up as point-sources, and will only appear on a few pixels in the image. Almost all the stars will appear the same. Galaxies, however, are large enough that we can see their shape, even though they are orders of magnitude further away than most stars.

And it's definitely nice to see work like this. I work on the star-side of SDSS, so I'm not as familiar with this particular classification project. But SDSS is pretty amazing overall. For some cool pictures and more details about the project, I'd recommend the movie (3min, very cool), some pie diagrams, and more pretty pictures.
posted by kiltedtaco at 8:58 AM on July 11, 2007 [1 favorite]


Ah good. Someone on the galactic side to give a more detailed explanation! Now I have to look up some of these papers.
posted by kiltedtaco at 9:00 AM on July 11, 2007


It got really popular once they added social networking features and a ratings system. Lots of people joined, but "EveryTed2007" was one of the first. Before long, he was one of the highest rated galaxy classifiers in the system. His accuracy - as measured by spot check against professional opinion, and the community - was nearly unrivaled, but the sheer volume of his work was what really set him apart. He assigned twice as many classifications as his closest competition. Part of this was that unlike many participants, he wouldn't skip the really difficult cases, but made extensive use of the custom "other" option to describe the galaxies not quite fitting into the standard choices in the dropdown. He'd write very specific - and surprisingly professional - descriptions for these cases, and before long the admins were adding his more commonly used phrases to the stock options.

Until the end.

After he stopped classifying, and suddenly seemed to disappear from the system entirely, they went back over his more recent entries. Everything was normal - his usual good work - and they picked up several new category names to add to the system. Except his last classification, which they could not get past the managers:

"Home."
posted by freebird at 9:13 AM on July 11, 2007 [8 favorites]


awesome post.
posted by afu at 10:23 AM on July 11, 2007 [1 favorite]


Sweet, I already found these strange galaxies.
posted by afu at 10:40 AM on July 11, 2007 [1 favorite]


This is ridiculously addictive, so much for getting any work done today!

Great post, thanks for this.
posted by jennaratrix at 11:00 AM on July 11, 2007


Looks like the two bluish galaxies are in the process of merging, afu. Nice find. The red fellow above them is a little bit* further away. This merging one has a redshift of 0.070 (just under a billion light years away) but this is at 0.089 - another couple of hundred million light years beyond that.

More mergers here.

*by 'a little bit' I mean comparatively speaking of course!
posted by edd at 11:11 AM on July 11, 2007


So if just need to check the redshift to compare how far away galaxies are (The Z value on the sky server pages)?
posted by afu at 11:34 AM on July 11, 2007 [1 favorite]


Got to be careful - if the object has a spectrum (and most things don't), then you can check the z value for that. But don't look at the z value for the photometry (that gets listed with u, g, r and i) as that's the apparent brightness through the z filter, not the redshift.

Fortunately the magnitude of an object in the z band is going to be a number in the ballpark of 20, and the redshift will always be a number less than 1 for a galaxy (and less than 7 for quasars), so it's easy to check you have the right one. There are of course galaxies beyond redshifts of 1, but SDSS won't have see them - they're basically too faint.

Ned Wright has a fantastic cosmology calculator which defaults to sensible cosmological values, so you can plug in a redshift and get out a distance (comoving radial is a good choice) or lookback time (use the 'flat' button).
posted by edd at 12:05 PM on July 11, 2007


This is fantastic! Mine are loading s-l-o-w-l-y, but that's fine. Just wanted to make sure everyone knows that if you don't register, it doesn't record your classification!
posted by misha at 12:52 PM on July 11, 2007 [1 favorite]


How weird is weird? I guess this is just a bad photo?
posted by hindmost at 1:33 PM on July 11, 2007


It's bad object detection resulting from a bright star. I've overlaid the sources that the software has picked out, and you can see it's picked out strings along the four diffraction spikes. As kiltedtaco said, stars look small. Which is usually true, but that one's just flooded the image.
posted by edd at 1:47 PM on July 11, 2007


Now it's working fine for me. I'm enjoying this a ton.
posted by Kattullus at 1:59 PM on July 11, 2007


Hey! It's The Great A'tuin!
posted by Kattullus at 2:39 PM on July 11, 2007


You know what would be really helpful? If after you took the test and passed, they identified the ones you missed. That kind of feedback would help fine tune the edge cases. I know it says not to agonize but I can't help it...
posted by hindmost at 5:16 PM on July 11, 2007


This is the most awsomest metafilter thing ever. I ~love~ astronomy and this helps me to realize my dream of being an astronomer, albeit indirectly and without learning any physics.

Now I find myself agonizing over the "tricky" ones, trying to figure out if its elliptical, spiral or should I mark it as something else? AHAHHHHHHHHHHHHHHHHHHHHHHHH
posted by Avenger at 5:35 PM on July 11, 2007


Ummm....WTF is this?!?

christ, I've been doing this for 5 minutes and I already found a massive anti-matter contrail from some giant alien spacecraft. sheeet...
posted by Avenger at 5:54 PM on July 11, 2007


full image here.

Maybe its a visible superstring or something?
posted by Avenger at 5:57 PM on July 11, 2007


gravitational lensing, perhaps?
posted by Avenger at 6:18 PM on July 11, 2007


I don't know, but I think it is expected to hit Earth on 1/18/08.

Actually, that's seriously cool, Avenger.
posted by misha at 6:19 PM on July 11, 2007 [1 favorite]


i can't begin to explain how awesome i find this. every time i look at a picture of agalaxy these days my mind gets warped by the scale of these things. now i'm looking at one after another that are so tiny they're like 30 pixels across and i might be the first person to ever see them???

*head explodes*
posted by garethspor at 10:13 PM on July 11, 2007


whoops, my bad, turns out its a satellite trail. (scroll to the bottom)

/still wishin it wuz alienz
posted by Avenger at 11:39 PM on July 11, 2007


With the gravitational lensing one it's sometimes hard to say. Even if the objects had spectra, which only one does, you'd still follow it up with observations from a bigger telescope to check. When looking for lensed images you have to deal with the fact that stuff out there is clustered, so you'll often see two very similar objects next to each other anyway.

I'd bet it is just that sort of clustering. We can use the colour of the galaxy to estimate the redshift - this is the 'PhotoZ' thing on the sidebar. One has a photo-z of 0.20-0.22 and the other is 0.23. The former is well within the expected photo-z error of the bottomost galaxy with a spectrum - and that has a definite redshift of 0.23, and obviously the one with the photo-z of 0.23 fits right in there too. So it looks like those three objects are all real-life neighbours (that guy to the north - not sure offhand what that is, could be a star sitting in the way or something).

Of course, you might say that if it were a gravitational lens then they could all be the same object so they'd all look the same, but you'd easily see any galaxies that could do the lensing in the middle, and there'd have to be a very big group to do it, and they're just not there.

So, my bet is it's an example of a galaxy cluster - albeit less visually impressive than the ones on that page.
posted by edd at 12:46 AM on July 12, 2007


kiltedtaco: But I don't mean to presume to take that criticism too seriously.

Nah, I'm just poking fun. Seriously, it's a worthy project, which is why I had a whole bunch of questions about it. (edd: thanks for taking the time to answer).
posted by Leon at 12:52 AM on July 12, 2007


This has been incredible. I've been coopted to help reply to the vast swathes of emails coming in to the project - thousands literally. A storming success so thanks to everyone who is classifying away!
posted by edd at 4:58 AM on July 12, 2007


« Older █ Pink Floyd The Wall [more behind]...  |  Shelfari. Books go Web 2.0.... Newer »


This thread has been archived and is closed to new comments