Planet Four Talk

Is there a doctor in the house?

  • AUricle by AUricle

    I'm in the middle of marking an image VERY similar to this--> http://www.planetfour.org/subjects/standard/50e73fa25e2ed212400025c9.jpg ...but the image is now frozen, in fact, the whole page is frozen after about 155 marks.

    I've got 32 unmarked features to go to FINISH this image. I'd also like to save this image to my PC for some 'play-time' browsing, but being frozen, I'm totally dead in the water. ( I was only able to get here by opening a new browser window)

    So what I'd like to know is this....Can anyone at P4 help me "unfreeze" this image so I can finish what was a time consuming and tedious job? I can leave the window open as long as need be.....

    This is not the first time I've run into the "too many features" problem, and I'm also wondering if there might be a way to 'cut up' busy images like this into 2, 3, or 4 sub-images, so that people can work around the limitations of the software or whatever the problem may be??

    I'm neither a scientist nor a data analyst, but I can add 2 + 2, and if your average 'value' for images includes many intentional 'miscounts', or skipped images, your average is going to be skewed. Maybe the scope of the project overall is so much greater than this one problem, that it makes little difference. I understand that an undertaking like this is bound to have some problem areas, so maybe I'm just shouting down a well. Hoping for some team response. Thanks in advance 😃

    Posted

  • Chocstar by Chocstar

    I feel your pain AUricle.
    I also lost a fabulous image with many features during a crash and never had a chance to save it or finish. It didn't show in my recent viewings either alas.

    Posted

  • AUricle by AUricle in response to Chocstar's comment.

    Hey Choc,

    Total subject switch...I notice you not only have an 'avatar', but you switched it, too!

    How do you get one here in the first place???

    Posted

  • Chocstar by Chocstar in response to AUricle's comment.

    Click on "profile" at the top of the screen. Click on the planetfour symbol silver circle on black square on the left hand side, (should say change your avatar). This should bring up the zooniverse website in a new window.

    Login with your same login details as this site. It should bring up your account setting, find the "avatar" tab and choose "browse" to find any pic you like on your computer. Afterwards press "update" you might have to do it twice. If it shows your piccie at the side then it should work for all the sites you log into connected with zooniverse. To change it just keep browsing and updating 😃

    Posted

  • SEJones by SEJones

    hi Auricle, may i offer help with your query about the stats around images? the mars project depends on a very large sample of human observers giving classification data. with a large classification dataset, project scientists can have confidence in the patterns that emerge. of course not everyone will agree on same feature - there will be some differences between observers but this is OK as long as you have a large sample size. but if a huge number of participants skip an image, this means that data is missing from many observers and results are skewed in favour of a much smaller sample that provided classification. the patterns from a very small sample is likely to differ from patterns based on a bigger sample. when collecting scientific data you want to avoid sampliing bias as much as possible. the more people involved the less variability in data and more confidence in results

    Posted

  • RobZTobor by RobZTobor

    Sadly I find if I get to 10 features without it crashing I am lucky, at this rate I will have to just give up as I cant do a decent job even if I try

    Posted

  • AUricle by AUricle in response to RobZTobor's comment.

    Hang in there Robot, .......there is supposed to be a fix in the works.

    Posted

  • AUricle by AUricle in response to Chocstar's comment.

    I knew I should've just kept quiet about the 200 object image I complained about 😉

    The gods are now having their way with me, as in this; http://www.planetfour.org/subjects/standard/50e73e0d5e2ed211dc001294.jpg
    I knew I had ZERO chance of marking this without the page crashing, even if I felt up to the challenge. So this is what I did.

    First I zoomed the browser window up to 400%. Next I took the 'blotch' tool, starting in the top left corner, made a daisy chain of blotch marks (at their smallest) like this--> oooooooooooooooooooooooooooooooooooooo, avoiding gaps or overlaps, ACROSS the image, until I reached the far right corner. There were 43 circle marks in this top 'row'. (each representing 1 column)

    Next, going back my 'start' mark in the top left, I did the same chain of marks DOWN the left edge, using the same 'rules', until I got to the lower left edge corner. The column had 32 marks. (each representing 1 row)

    OK, so that was enough info to know that it would take at least 1376 blotch marks to entirely cover the image ( 43 columns x 32 rows = 1376 marks) and in polling the 75 blotch mark circles I made, I found each circle had an average of 3 blotches in it (some as many as 6, a few with 0, most with 2-5) so 1376 marks x 3 blotches/mark = 4128 individual blotches....and that is a 'minimum' figure. It is NOT an 'exact' figure, but a good 'average', I believe. Also of limited if any value to the scientists since it aggregates blotches

    The caveat here is that this counting method only works (with any accuracy) for very homogeneous images such as this one. How many images are like this I have no idea, but I know I've seen at least 5. So maybe there are only 5, but maybe there are 1000.

    Posted

  • mschwamb by mschwamb scientist, translator

    Hiya,
    Just a quick note, but I wanted to say that we're aware of this crashing/freezing bug, and a long term fix will be implemented by the Zooniverse development team. I don't exactly know when or if it's been implemented yet, but the team is definitely working on it.

    Cheers,

    ~Meg

    Posted