Chicago Wildlife Watch Talk

Some info about how this works would be nice

  • buffalogroveP by buffalogroveP

    Hi - I've noticed that several of us have viewed/ classified the same photos -- could one of you people in charge explain how this is all working? i really think that I've seen the same photo more than once. I realize that maybe it's all randomized, and it is also good to hav more than one opinion on a pic.... but an explantation about how it's all supposed to work would be helpful and nice. thanks 😉

    Posted

  • mason_UWI by mason_UWI scientist

    My understanding of the system here is that a general consensus needs to be made on photographs before it actually is considered identified. Some photos may look to be the exact same, but there are times when we get a lot of similar looking photos caused by a stick blowing in the wind. On top of that, you may get the same photo every once in a while to verify your previous tags. The zooniverse people would be able to elaborate a fair bit more.

    Posted

  • DZM by DZM admin

    Yes, we want to get many people's opinions on each pic! If we get 10 people who all say that a picture is "two gray squirrels," then we can be fairly sure that's what it is. But if we get 2 people who say muskrat, 2 who say gray squirrel, 2 who say fox squirrel, 2 who say fox, 1 who says opossum and 1 rebel from Snapshot Serengeti who says giraffe, we know that that photo should probably be analyzed by a pro! 😃 That's why even your best guess is always extremely helpful--if there are a bunch of wild guesses, that's also informative to the science team.

    In terms of seeing the same photo multiple times, I think that I've run across that as well. I think that @mfidino@lpzoo.org is definitely right about similar-looking photos, and about occasionally tripping over the same photo... Is it happening to you a lot?

    Thank you for the feedback!

    Posted

  • octopus.gallery by octopus.gallery

    There are certain sites where there is definitely vegetation that's triggering the camera (the canal is one that comes to mind, or shots of the tree canopy) because you can see it in the shot each time in a different position. Others I suspect are just critters moving through at speed like squirrels or birds. There was one site that I kept seeing empty multiple times and I could tell they were all within a short time frame or were all the same picture because there was a dandelion that had gone to seed in the foreground and it always had the same amount of seed on it. Those don't last long.

    Posted

  • buffalogroveP by buffalogroveP

    Thanks for the info. I was suspecting that it was a consensus thing. And i was worried about making wild guesses -- so thank you this helped a lot.... I won't be afraid to make a wild guess...

    Posted

  • DZM by DZM admin

    Good! Yes, please, make a wild guess! We're working on making our introductions to the projects all say something along the lines of "don't be afraid to make a wild guess." 😃

    Posted

  • escholzia by escholzia

    I'd like to know if this project tracks classified images for each participant. I understand that there are many, many similar images per camera, but I hate to think I'm classifying the same pics over and over again. If there are 1 million images, as stated on the project page, and I classify 10,000 of them at random (not there yet, but it does get addictive), by probability theory I would expect to see 1% repeats of any given image if I keep seeing already classified images. I'm not math nerd enough to say how often I'd see repeats of any image in 10,000 classifications, but I think it must be very high percentage. At some point, it becomes diminishing returns. So.... could you find this out for us?

    Posted

  • mason_UWI by mason_UWI scientist

    Hi! While there are 1 million images, we only classify one season at a time (takes the # of photos uploaded to around 100,000). On top of that, some photos are already considered 'classified' (42%). So there are around 60,000 photos that need to be classified. Of those photos, a decent chunk have nothing in them. So if a camera has collected a large number of photos with nothing you may think that you are classifying the same exact image, but it may be an different photo all together. Additionally, from the data I have entered on the site (yes it is addicting) it seems that the photos get pulled up in chunks (e.g. grab a random photo from these 4-6 sites), and once those are complete move to the next few sites.

    So there is a chance that you have entered a photo multiple times, but it is also likely that you have entered multiple photos that look similar. Having a single photo entered multiple times by one person is not bad though, as it allows us to verify if the user is consistent with their photo identification.

    So why do we do one season at a time? From a programming stand point it is likely less server space (which can add up), but on the scientific end of things (what I do) its a better way for us to be able to get data that we can start working with, which will then allow us to start showing all of you what we have found out from all the identifying everyone has done.

    I hope that answers your questions!

    -Mason

    Posted

  • escholzia by escholzia

    Thanks, that helps a lot. I like to know that my efforts are productive. Sometimes I feel like the kid in The Phantom Tollbooth, moving grains of sand from one pile to the next. I'll figure out at what point (statistically) the repeats will get excessive and stop until the next data set is loaded.

    I'm sure my repeat identifications are not always consistent, since I only discovered the blog posts after I'd already done a few hundred classifications. They are very helpful - you should find a way to link them to the home page!

    Posted

  • mason_UWI by mason_UWI scientist

    @escholzia that would be a great idea! @DZM anyway for the ID blogs to be more easily available?

    Posted

  • DZM by DZM admin

    We could probably do something! Do you know who built your site? Scanning it right now, I can't find any easy or obvious links to the blog. We could probably add one wherever you think that would be proper.

    Send me a PM or email (darren@zooniverse.org) and we can hash it out!

    Posted

  • nancybohrer by nancybohrer

    I think there needs to be an area with more pictures of the animals. I don't see woodchucks, beavers, muskrats or foxes in daily life, and if you want better data you need to provide better information. Also, auto/other is very unclear.

    Posted

  • DZM by DZM admin

    Hi @nancybohrer , "vehicle/other" is just for a car or any animal that isn't listed (say, if the camera caught a glimpse of that cougar that came down into Chicago a few years back). Hope this helps!

    As for the images, that's not a bad idea. We do indeed do something like that on Snapshot Serengeti. For now, maybe Google Images can help?

    Woodchucks

    Beavers

    Muskrats

    Foxes

    Thanks!!

    Posted

  • escholzia by escholzia

    Thank you! The link beneath the photos is very visible, and the blog posts are very helpful.

    Posted

  • DZM by DZM admin

    Glad you noticed it, @escholzia ! It was really good feedback and something that I'm actually trying to implement on as many projects as possible, especially new ones... having something in the classification interface for "frequently asked questions." 😃

    Posted