Radio Galaxy Zoo Talk

ARG00000eb - yet another zoomed out, possibly mismatched image?

  • JeanTate by JeanTate

    Once again, this field does not look like the center of the larger field that I just finished classifying (screenshot etc to follow). The WISE images seem to match the IR sources; the FIRST and NVSS images also seem to match the radio sources ... but neither seem to match the sources - of either kind - in the field I was offered to classify. For example, there was no nice bent doublelobe ...

    Posted

  • JeanTate by JeanTate in response to JeanTate's comment.

    Here's the 'Finish' screenshot:

    enter image description here

    And here's the central third, 'blown up' to the same scale (±a few percent) as the image of ARG00000eb to the left:

    enter image description here

    Posted

  • DocR by DocR scientist

    Thanks for the documentation. I'll pass this on to the software people.

    Posted

  • edpaget by edpaget

    @JeanTate, So it looks like you were served data from the Beta test (which had larger fields like the one you posted). Can you double check you haven't been going to radio.galaxyzoo.org/beta/ or radio.galaxyzoo.org/beta2/ (which I will take down today anyway)?

    Posted

  • JeanTate by JeanTate in response to edpaget's comment.

    In my page1 "Recents" there are 12 images; in reverse chronological order (i.e. most recent backwards):

    I do not recall what the URL of the 'classify' page I was on was; however, the transitions for most - possibly all - the above (i.e. from one classify image to the next) were triggered by me clicking on the "Next" button.

    it looks like you were served data from the Beta test (which had larger fields like the one you posted).

    Even if that were so, why would the fields seem to be centered on (quite) different (RA, Dec) locations? For example, it seems to me that none of the larger fields contain radio sources displayed in the "ARG..." fields, anywhere ...

    And I would say that, among zooites, I'm particularly vigilant about this sort of thing; if it's happened to me - at least three times - how many times has it happened to other zooites, in the last two+ days, without anyone posting, or even "Discuss"ing?

    Posted

  • edpaget by edpaget in response to JeanTate's comment.

    The fields presented in the current version of Radio Galaxy Zoo are very different than the ones shown in the beta version, both in scaling and the surveys they were taken from.

    However, since the two datasets live on different servers they can have the same "ARG..." field, which means that it will take you to this page if you click the discuss button.

    At least I think that's what's going on based on your description. Let me just try to be clear, you were classifying and saw the top picture from your post, then clicked discuss, and got taken to this page and saw the different image there?

    Posted

  • JeanTate by JeanTate in response to edpaget's comment.

    Not sure I understand: the ARG0000xxx ids are not unique? It's possible for the same id to refer to two different fields, with different (RA, Dec) center-locations, different scales, different source data (surveys)?

    Also, I got the same field to classify, twice; according to Arfon, this is impossible in the current Zooniverse design.

    you were classifying and saw the top picture from your post, then clicked discuss, and got taken to this page and saw the different image there?

    Yes.

    And it happened to me five times in a row, with three different images.

    And it seems to have happened to firejuggler too (second post).

    Posted

  • edpaget by edpaget in response to JeanTate's comment.

    Yeah we host subjects for beta testing on a different server.

    The identifiers are simply a three letters designating the project (A Radio Galaxy Zoo), and then the subjects are numbered sequentially in base32 (or 36, I'm not totally sure off the top of my head).

    When we load our subjects onto the beta server, they are numbered starting with ARG000001, then ARG000002, etc, and when subjects are loaded into the production server the first is ARG000001, the next is ARG000002. They will be unique for all subjects loaded onto a specific server, but not necessarily unique between the main version and the beta version.

    As far as subjects happening twice, there are only a few subjects on the beta server, and they have no such restriction about being served more than once, (so it doesn't tell me I've classified everything while I'm developing the site).

    Were you a beta tester for this project? I think I've resolved any way I can think of for you to be served beta subjects, but if you see any more like this please post here, or send me a message about it.

    Posted

  • JeanTate by JeanTate in response to edpaget's comment.

    Thanks.

    Were you a beta tester for this project?

    Yes, I was - formally - a beta tester (though in the end I did precious little; mostly it either froze/crashed, or I couldn't even get started/beyond the Tutorial!).

    if you see any more like this please post here, or send me a message about it.

    Will do. By Talk PM, or ...?

    Posted

  • edpaget by edpaget in response to JeanTate's comment.

    Talk PM would be great.

    Posted

  • JeanTate by JeanTate in response to edpaget's comment.

    Thanks; will do.

    Posted

  • JeanTate by JeanTate in response to edpaget's comment.

    Yeah we host subjects for beta testing on a different server.

    The identifiers are simply a three letters designating the project (A Radio Galaxy Zoo), and then the subjects are numbered sequentially in base32 (or 36, I'm not totally sure off the top of my head).

    When we load our subjects onto the beta server, they are numbered starting with ARG000001, then ARG000002, etc, and when subjects are loaded into the production server the first is ARG000001, the next is ARG000002. They will be unique for all subjects loaded onto a specific server, but not necessarily unique between the main version and the beta version.

    It's been a long time since I was involved in databases, and I've surely forgotten far far more than I care to admit, but reading this caused all kinds of alarm bells to ring in my head. In one of the IRL jobs I had, when someone else - in a different department (fortunately) - did something like this it created no end of hassles, messes, and trouble (not to mention lost customers). Not only was it a key SOP that database IDs absolutely, positively HAD to be unique - for a bazillion 'good database management/practice' reasons - but the poor guy who messed up did so in much the same way1 ... he assumed that because the databases were held quite separately, it was OK to create non-unique IDs. Of course, someone else had also - independently - broken a different SOP by skipping a step or two, creating a situation where access to the (supposedly separate) databases could 'cross over'. Customers who were able to access data on different customers were not pleased.

    In both the GZ forum and GZ Talk I started threads on the broader topic: How is the data integrity of the Galaxy Zoo 'clicks databases' assured? (in Talk), How is the data integrity of the Galaxy Zoo 'clicks databases' assured? (in the forum). Would you be able to write a post or two on this topic, in either thread? Thanks in advance.

    In cleaning up this, were you able to identify the clicks (classifications) which referred to the beta Subjects (not just mine, but those of any and all other zooites who may have also accidentally classified beta Subjects)? And were you able to be 100% sure that no 'production server' classification clicks were lost, by ending up in the beta database by mistake?

    1 Or not; it's a long time ago, and I'm sure I've forgotten some important details

    Posted

  • HelPer2 by HelPer2

    Just a quick comment - I've personally only seen the "new" faulty type images linked to the wrong "Discuss" page in the past few minutes, although they have repeated (The one initially reported by Jean above was served to me twice in a few seconds)

    Posted