Author Topic: Why we need you NOW  (Read 25305 times)

Edd

  • Admin
  • Hero Member
  • *****
  • Posts: 2638
    • View Profile
Re: Why we need you NOW
« Reply #15 on: November 30, 2007, 02:21:05 am »
It's not easy to work around that issue, Johnny - I think Chris posted on it elsewhere. Still, it is a useful thing to do as if the  results do change we know there is a bias there, even if it's still difficult to assess the extent of it.

Anyway, it's a known issue, and at the end of the day a compromise had to be made.
When I look up at the night sky and think about the billions of stars out there, I think to myself: I'm amazing. - Peter Serafinowicz

Hanny

  • Global Moderator
  • Hero Member
  • ****
  • Posts: 17256
  • "Voorwerp kid"
    • View Profile
Re: Why we need you NOW
« Reply #16 on: November 30, 2007, 05:22:58 am »
Chris DID say something about that.. probably in the link Fluffy posted. Yes, you often see that pics are changed somehow (rotated, B&W..) and yes, you can still have a look at the sdss page, BUT it is made clear, from the beginning, that you should only classify from what you SEE and that hasn't changed. So, please go on classifying those strange pics just by looking at them. :)

Rick Nowell

  • Hero Member
  • *****
  • Posts: 1333
  • Bristol, U.K.
    • View Profile
Re: Why we need you NOW
« Reply #17 on: November 30, 2007, 10:26:07 am »
Since the bias test images have been introduced, after doing a modest 750, i 'evolved' a method whereby i make a decision in the Analysis, which i stick with, before going into SDSS quasar hunting or whatever, noticing how the image looks as if it were normal.

Using that, i've concluded that the reversal or rotation matters little, to me, but the b&w's definitely have me thinking, or not acting
on auto-pilot, so to speak. Images that are rotated have the smallest effect, as who knows how they are anyway?

They do make me pause though, which is probably a good thing. The arcane world of statistics.

EDIT:  This one was alot easier to classify in its 'normal' image in SDSS than when in the Analysis, where it has been rotated. It's
difficult to say that this is because of a subconcious bias, or just because it was harder that way round anyway as an image per se.

587742901795881205
 
« Last Edit: November 30, 2007, 03:59:38 pm by Rick Nowell »

pluk

  • Hero Member
  • *****
  • Posts: 7161
  • serving coffee on aircraft causes turbulance
    • View Profile
Re: Why we need you NOW
« Reply #18 on: November 30, 2007, 02:30:19 pm »
It's not easy to work around that issue, Johnny - I think Chris posted on it elsewhere. Still, it is a useful thing to do as if the  results do change we know there is a bias there, even if it's still difficult to assess the extent of it.

Anyway, it's a known issue, and at the end of the day a compromise had to be made.

Did you consider installing a (sorta-kinda) proxy to the SDSS object explorer (and other tools), one that applies the exact same transformation as was used to produce the bias testing image (which I think is either flipping, or grayscaling or both) and just passes on everything else?

edit: All that would really be required is a version of the image-cutout-tool (http:/.../ImgCutoutDR6/getjpeg.aspx) with a few additional drawing options.
« Last Edit: November 30, 2007, 02:36:29 pm by pluk »
We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run. - Roy Amara, engineer, futurist (b. 1925)

Mark OConnell

  • Sr. Member
  • ****
  • Posts: 418
    • View Profile
Re: Why we need you NOW
« Reply #19 on: November 30, 2007, 05:46:56 pm »
Cool! I have already banged out like 50 or more. I will have more time over the weekend.. See if you can send out an email to all past classifiers to come and help with this next step.

*encouragement* Remember we as in ALL of us came here to help do science. So E-mail us ALL for help.

Johnny

  • Jr. Member
  • **
  • Posts: 90
    • View Profile
Re: Why we need you NOW
« Reply #20 on: December 01, 2007, 12:58:52 am »

Hey folks,

It just dawned on me that when The Keeper reverses the image of a spiral (i.e. twists it 180*) we, now, are looking at it as if from a point on the exact opposite side of the cosmos. 

So, why not do the same to the "edge on spirals" but only twist them a quarter of the way around (90*) and then we can determine if they're "clock/anti-" or just cool little pencil galaxies?

I think I've been looking at these things a long time now.

Johnny

NGC3314

  • Global Moderator
  • Hero Member
  • ****
  • Posts: 1590
    • View Profile
Re: Why we need you NOW
« Reply #21 on: December 01, 2007, 03:41:42 am »
With the B&W versions turning up now - keep in mind that this is how the classical systems of galaxy classification (Hubble, Sandage's extension, de Vaucouleurs, Morgan, Vorontsov-Velyaminov) were defined (albeit often working from the original negatives on glass), with color information coming more slowly and indirectly from comparison of photographs through different filters. Being able to even think about wholesale classification using color put together in a consistent way for so many galaxies is a new thing - so it will be interesting to see whether not starting with color makes a measurable difference. (As in, not being able to just click E because of that golden-bronze sheen). Indeed, one interesting aspect of GZ is seeing what kind of "natural" classification might emerge from observers not preconditioned to the families defined in these earlier systems. I might suggest GZ3 should throw in 2MASS and GALEX imagery to widen the spectral coverage, but I'm pretty sure that would result in the zookeepers dispatching some burly individuals to rough me up at an upcoming meeting because they've had to go through quite enough with the new features as it is. So forget I typed that..

Edd

  • Admin
  • Hero Member
  • *****
  • Posts: 2638
    • View Profile
Re: Why we need you NOW
« Reply #22 on: December 01, 2007, 10:42:53 am »

Hey folks,

It just dawned on me that when The Keeper reverses the image of a spiral (i.e. twists it 180*) we, now, are looking at it as if from a point on the exact opposite side of the cosmos. 


Twisting 180 degrees is not the same as a reflection (just to be pedantic)
When I look up at the night sky and think about the billions of stars out there, I think to myself: I'm amazing. - Peter Serafinowicz

ElisabethB

  • Hero Member
  • *****
  • Posts: 13200
    • View Profile
Re: Why we need you NOW
« Reply #23 on: December 01, 2007, 10:54:41 am »
that's our Edd !  ;D

fluffyporcupine

  • Global Moderator
  • Hero Member
  • ****
  • Posts: 8179
    • View Profile
Re: Why we need you NOW
« Reply #24 on: December 01, 2007, 10:56:34 am »

Hey folks,

It just dawned on me that when The Keeper reverses the image of a spiral (i.e. twists it 180*) we, now, are looking at it as if from a point on the exact opposite side of the cosmos. 


Twisting 180 degrees is not the same as a reflection (just to be pedantic)

surely it depends which axis you are twisting about ;) :D

njtm

  • Hero Member
  • *****
  • Posts: 1731
  • The Wonder Down Under :)
    • View Profile
Re: Why we need you NOW
« Reply #25 on: December 01, 2007, 11:03:11 am »
I think this thread is getting twisted ::)

fluffyporcupine

  • Global Moderator
  • Hero Member
  • ****
  • Posts: 8179
    • View Profile
Re: Why we need you NOW
« Reply #26 on: December 01, 2007, 12:16:08 pm »
Please everyone, if you find a lovely black-and-white merger, can you post it here for me?

here ya go Scary

587745244704276545

fluffyporcupine

  • Global Moderator
  • Hero Member
  • ****
  • Posts: 8179
    • View Profile
Re: Why we need you NOW
« Reply #27 on: December 01, 2007, 04:29:01 pm »
Did you consider installing a (sorta-kinda) proxy to the SDSS object explorer (and other tools), one that applies the exact same transformation as was used to produce the bias testing image (which I think is either flipping, or grayscaling or both) and just passes on everything else?

It would be really useful if could do this as i get very confused trying to work out how the image has been rotate for the really zoomed in pics (e.g. the NGC ones) to relate what comes up in analysis and what comes up in SDSS. with most of these you cant determine the classification properly without zooming out. unless the NGC ones could be excluded from teh mirroed part?

Johnny

  • Jr. Member
  • **
  • Posts: 90
    • View Profile
Re: Why we need you NOW
« Reply #28 on: December 01, 2007, 07:54:18 pm »

I've been noticing that I spontaneously respond a bit differently depending on where the spiral meets the galaxy's core.  For example, a galaxy with 2, relatively short, spiral arms that turn sharply to the core at 3:00 and 9:00 o'clock slow me down a lot.

I don't know if this is a feasible/useful bias check but if:

1.  one took a bunch of simple 2-armed spiral galaxies and first blacked out an inner perimeter around each one's core--leaving only the outer portion of the arms,
2.  then took the same photos and blacked out only their outer circles, obscuring the trailing tail of the spirals, and
3.  compared how people analyzed the 2 groups,

would one find a bias.





EricFDiaz

  • Guest
Re: Why we need you NOW
« Reply #29 on: December 01, 2007, 08:29:13 pm »
Did you consider installing a (sorta-kinda) proxy to the SDSS object explorer (and other tools), one that applies the exact same transformation as was used to produce the bias testing image (which I think is either flipping, or grayscaling or both) and just passes on everything else?

It would be really useful if could do this as i get very confused trying to work out how the image has been rotate for the really zoomed in pics (e.g. the NGC ones) to relate what comes up in analysis and what comes up in SDSS. with most of these you cant determine the classification properly without zooming out. unless the NGC ones could be excluded from teh mirroed part?

Fluffy, I wouldn't worry about the "really zoomed-in" pics for now. With only one exception, I haven't been going to SkyServer at all while this bias test is ongoing. If I get an image, color or B&W, that is monochromatic, I click Star/Don't Know. The only exception to which I referred above are S0 galaxies, because they are a kind of pet project of mine, and I don't want to miss out on potentially significant finds. But, I do this only after I have decided on a classification of S0. I don't feel that this affects the outcome of the testing for bias since S0 galaxies are classified as ellipticals anyway, and this "bias" testing is looking for bias in classifications of ACW vs. CW spirals. But, if I get a flat patch of gray as an image, I just classify it as Star/Don't Know, even if it may be a spiral that can be recognized by zooming out.That's just the way I do it.