Chicago Wildlife Watch Talk

Arthur by Numbers

  • ForestPreserve by ForestPreserve moderator

    I have found 5,858 images of Arthur over the past couple of months: CCHS000043

    Happily, I think this is all of them.

    Many of these images came up during classifying, but I was able to get the full collection by prospecting on the "talk" page. For example, the first Arthur I collected was ACH0000gyj. I realized that ACH0000gyi was also Arthur, as was ACH0000gyk, and an obsession was born.

    The interesting (?) part is that Arthur is found in one pretty solid block. It starts with ACH0000ec9, which has a squirrel bounding by, and ends with ACH0000jcy, just Arthur on his own.

    In this block, the only non-Arthur images seem to be "barriercam", for example: ACH0000ecc

    So far, all the talk pages I've seen are in the range ACH0000^^^ through ACH0002^^^, where the "^^^" positions can be '0' through '9' and lowercase 'a' through lowercase 'z', so 36 possibilities. The ACH0000^^^ range would have 36 x 36 x 36 = 46,656 images. The ACH0001^^^ ranges doubles this to 93,312 images.
    So far, the highest image seems to be ACH00023k8, so another 4,617 images there, for a total of 97,929 available.

    So Arthur is about 6% of the available images, 5,858 out of 97,929.

    If you consider greater Arthurville, ACH0000e^^ through ACH0000j^^, that's 6 * 36 * 36 = 7,776 images, 75% of them Arthur.

    In the town of Arthurville proper, ACH0000ec9 through ACH0000jcy, there are 6,039 images, 97% of them Arthur.

    The total size of 5.858 Arthur's is around 450 MB. I assume most cameras can store at least 1 GB of images, so it's possible all 5,858 Arthurs were collected in one placement of the camera.

    Based on the clustering of images based on their talk page URLs, I suspect that all the Arthur images were added to the database at one time. I'm guessing that images from the "barriercam" camera were being loaded at about the same time, since these are the only ones that intrude in Arthurville, and they seem to make up much or all of greater Arthurville.

    Hopefully, the Arthur cam has been adjusted or replaced and moved to a new location.

    Take a look at my mower collection CCHS00004e for an example of images being grouped by talk page URL, ACH00016dk and following.

    Posted

  • DZM by DZM admin

    This is one of the most fascinating things I have yet seen accomplished by a zooite.

    No, I'm serious! This actually illustrates perfectly just how much the project can be affected by something as silly as a dumb plant flapping next to the camera. 6% of the available images is a significant number, to the degree where Arthur has probably actually been costing us a fair amount of efficiency on this project.

    Thanks so much for the really impressive analysis, @ForestPreserve ... I'd love to hear some of the scientists' thoughts on if there's anything we can do to banish Arthur straight away, and if ArthurCam has indeed been permanently moved... and if there's a lesson in image collection to be taken from this? 😃

    Posted

  • mason_UWI by mason_UWI scientist

    Typically, when we get an issue like this with a camera we need to determine if the malfunction (i.e. the fact that the camera took 4000+ photos of a plant) was arthur's fault or the cameras fault. Some of our cameras are a couple years old, so when we do get an "arthur" we take those cameras and test them in the office to see if the camera is overly sensitive. If it is, we retire it (we have lots of photos of shoes in the hallways). If the camera performs well though, it goes back into circulation. We do our best to properly place cameras, but given the fact that there are >110 sites there tends to be at least 1 that gets some swinging vegetation and 1 or more sites with cameras that need to be retired.

    In the future we hope to prevent future arthurs with some additional screening before the photos get uploaded so we can keep large blocks of "nothings" from happening. There's no immediate way to remove them on our end, as none of us have any skills with the programmatic analysis of images, but we should be able to decrease the arthurs with some pre-screening.

    Posted

  • escholzia by escholzia

    Wow, ForestPreserve - I can't believe you collected all those images! I bow to your obsessiveness...

    mfidino - Do you collect metadata along with the picture files, such as camera ID and time/date stamp? If so, it should be fairly easy to do a quick plot of number of pics for each camera over time. That will pick up frequency spikes that are possible malfunctioning cameras or arthurs. Those outlier pics can be confirmed then deleted before inflicting them on your loyal classifiers.

    btw- I'm really looking forward to the next batch of photos. Fall leaves are great, but show us some snow.... Time to start the prescreening now - we're at 86% on the current batch!

    Posted

  • mason_UWI by mason_UWI scientist

    We're already compiling a list of sites to take out. And yes we collect metadata (and I'm currently in the process of programmaticly collecting that info from the camera photos). For now, we are just going to put sites to the "side" for us to enter if they have >2000+ photos.

    Thanks for helping us through the first season everyone, hopefully the next one will be full of more wildlife and less arthurs!

    -Mason

    Posted

  • DZM by DZM admin

    Thanks, Mason!

    How close are we to moving on to "Season 2," out of curiosity?

    Posted

  • juliamk by juliamk

    Aha, I finally figured out who Arthur is! I've been on here for a couple weeks and had seen some references, and the tag, but not this discussion. I had thought perhaps Arthur was someone's name for the invisible being who was walking by cameras and tripping the motion detector. Now I see he's one very photogenic plant. Nice to meet you, Arthur. 😃

    Also, I second the desire for snow pics. 😃

    Posted

  • mason_UWI by mason_UWI scientist

    It all depends on when this season get's finished!

    I'm very excited to start making maps showing "observed" distributions throughout the city!

    Posted

  • DZM by DZM admin

    Oooh, maps would be wonderful!

    Also, Mason, a question: Once the upcoming season of Snapshot Serengeti is finished, I'm considering sending them over here to CWW. It's likely that you'd get a huge flood (Snapshot's user base is huge) and get your data classified extremely quickly. But I'm not sure if you want that or prefer the more deliberate, growing pace that we have right now.

    Thoughts?

    Posted

  • escholzia by escholzia

    Any chance you could create a Google map showing the camera locations? That would be so cool if classifiers could relate what they're seeing to the environment the camera is in. Although on second thought, you wouldn't want to make the map TOO accurate - there's a potential for misuse.

    Posted

  • DZM by DZM admin

    you wouldn't want to make the map TOO accurate - there's a potential for misuse.

    Yes, this is one reason why they generally prefer not to discuss their cameras' locations. Although there's probably less danger of encouraging any sort of poaching in Chicago. But still perhaps some.

    Posted

  • mason_UWI by mason_UWI scientist

    Yeah, we will plot out distributions at broad scales (with a bit of error) to keep the actual locations from being found. Also, because we are interested in human use of sites we don't want to encourage CWWers to go on a scavenger hunt for our cameras!

    Also, DZM, please encourage any and all to help us enter camera trap data!

    Posted

  • DZM by DZM admin in response to mfidino@lpzoo.org's comment.

    please encourage any and all to help us enter camera trap data!

    Alright, you asked for it. 😃 Serengeti is about to get new data, but they process it with incredible alacrity. When they run out, I'll send 'em here... but you had better be ready as much data as possible, or else we might run out here, too!

    Posted

  • nancybohrer by nancybohrer

    I had no idea there are that many sites. Maybe you can assemble an increased group of animal id's for next season which would be very helpful, and it sure would help to have a reference shot of a site where we could see some measurements.

    Posted