Chicago Wildlife Watch Talk

Image ACH000d5cc / Drawing Tool?

  • WillowSkye by WillowSkye

    Deer next to tree

    I really hope this works and that you see the picture (first time I'm trying this)

    I suck at directions. Do you also see legs, chest, head & antlers? Otherwise I have been looking at too many photos ...

    With understanding that the classifications have already passed the 60 % marker, would it be possible, if time permitted, to create a "freehand" drawing tool that appears next to "DISCUSS" which allows one to draw a red circle on the photo where sometimes one sees something in the background which is either very small/faint that resembles an animal? Instead of giving you weird directions, at which I suck anyway, like (not applicable to above photo) "in the middle of the photo is a big, upright tree, go to the right of that until you see a little, upright tree, go to the right of that again until you see a middle-sized upright tree. 2 fingers' (horizontal) width from where the tree meets the ground, along the trunk, just to the left you'll see an opossum's head sticking out, sniffing at the ground (by this time the 140 characters have long since been used up). With the tool, one can just draw the red circle around what looks like the head and write "#opposum?"

    If it is crap, you just dump it in the Recycle Bin, but if it is right, you gain an animal. As distance from the camera increases, it gets more difficult to see similar things in the background. Like the person who saw a cow awhile ago and others a log. For the life of me, I still cannot see the cow, but if the person had marked the area on the photo, one could see what the person saw/meant. Photos marked as "nothing" might just have something. Like a cow...

    Posted

  • mason_UWI by mason_UWI scientist

    That would be a really cool idea I think!

    Posted

  • WillowSkye by WillowSkye

    Thanks 😃

    Posted

  • ForestPreserve by ForestPreserve moderator

    WillowSkye, amazing that you can spot things like this. The animal is gone in ACH000d5cd. I set up a "blink comparator" with ACH000d5cc and ACH000d5cd (just two tabs with the images aligned the same way). It looks like a deer facing away from the camera and to the left. I think I can see antlers.

    Check out the Wildebeest Watch, which is currently in beta test. Apparently zooniverse sign-in is not working for Wildebeest Watch, but it does have several interesting features.

    When you spot a wildebeest (don't worry, there are plenty), you get to specify which direction it's moving. You select a direction, then click on a wildebeest. This draws a circle with crosshairs on the wildebeest (don't worry, we're only shooting...pictures). You repeat this process for each wildebeest you can see. Sometimes you see a lot of wildebeest.

    In addition to the drawing tool, and you can play a series of images and pick the image you want to identify. There's a play/pause button on the lower left of the image which flips through the images in sequence. You can also flip through the images manually by clicking the circles to the right of play/pause. (Other projects have a similar play feature.)

    The circle-drawing feature could be the basis of an enhanced commenting tool. (Not sure how much effort would be required to retrieve and redraw the circles, however.) The playback feature would also be nice in resolving some of the tough cases.

    That being said, I suspect the tough cases like this one aren't the main scientific focus of the Chicago Wildlife Watch. (I'll let the scientists chime in on that.) I think the tough cases help us figure out what's what.

    Posted

  • WillowSkye by WillowSkye

    Thanks again 😃 Had a look at Wildebeest Watch and there is a lot of Wildebeest! The project is almost like a synthesis between Penguin Watch and Snapshot Serengeti with the difference being that one is only looking at Wildebeest. It becomes difficult when there are a lot of them, similar to Penguin Watch, and the further away they are from the camera. The options "Away from camera" and "Lying down, not moving" have a similar colour blue which causes a bit of confusion. Good news is, you can sign in with your Username and Password.

    Overall, looks like a great project, but I like this one better as I am seeing nature on another continent which is totally different to where I am. One tends to become overly familiarized with animals in your own "back-yard" even though they are elephants, lions, wildebeest, etc. Not that they are physically in my back-yard. They live in game ranches, game/nature reserves, national parks and zoos. However, a very clever young, 3 year-old lion did escape from the Karoo National Park about 3 weeks ago and is still on foot among the farming communities. It has killed about 24 sheep so far, with one ewe weighing 70 kg. It has also killed a kudu of about 150 kg. The lion seems to be having a great time and is so clever that it has even eluded professional trackers brought in from neighbouring countries.

    Posted

  • WillowSkye by WillowSkye

    ForestPreserve, I have reread your previous post and realized that I do not think I really understand your last sentence

    "That being said, I suspect the tough cases like this one aren't the main scientific focus of the Chicago Wildlife Watch. (I'll let the scientists chime in on that.) I think the tough cases help us figure out what's what."

    Would you please explain it to me? Furthermore, how does the above relate to Wildebeest Watch?

    I sometimes misunderstand what is being said as English is unfortunately not my First Language. It is my Second Language.

    Posted

  • ForestPreserve by ForestPreserve moderator

    WillowSkye, not sure this was the best example of English prose. I was thinking that we spend a lot of time on the tough cases. I enjoy it, but is it helping with the scientific goals of the project? Or would it be better to spend more time classifying the easier cases?

    Since then, I found Mason's "are we getting it right?" blog post. They checked a sample of 1,200 images, and found about 10% had incorrectly been classified as "nothing here". If I understand the blog post, the 1,200 images checked were ones which showed disagreement in the classification. So I think there's a strong case for spending a little more looking at the images, including figuring out the tough cases.

    I brought up Wildebeest Watch because it shows what's technologically feasible. The circle-drawing identification tool could be the basis of an enhanced talk tool like the one you describe.

    And speaking of lions on the loose, there was the case of the Chicago cougar a few years ago.

    Posted

  • mason_UWI by mason_UWI scientist

    Hi all,

    Yes, in way every photo is important, but we realize that some things are going to fall through the cracks and that some photos you just cannot ID. All of this is what we would call part of the observational process (in other words, how we get our data). However, when we will be creating statistical models for these data, we also will be inherently accounting for any potential bias that could be introduced, because clearly there is no way to be 100% certain that a camera trap photographed the entire wildlife community, it was active for a decent number of days, and then it was 100% correctly identified on Chicago Wildlife Watch. These types of models are called 'hierarchical' because the latent process we want to understand (species presence/absence) is linked to things like habitat type, but it also happens to be linked to the observational process (i.e. we have data that influences different parts of the whole problem and can account for introduced bias).

    That being said, more accurate data is always better! This is before my first cup of coffee, so hopefully it is coherent enough for everyone to understand!

    -Mason

    Posted

  • WillowSkye by WillowSkye

    Thanks, ForestPreserve & Mason 😃 I think I get it. It is similar to what is taught in Geology i.e. "Don't get carried away by big, beautiful crystals/minerals in one small area - take more samples over a larger area so that you will be able to get a better idea of the lay of the land." Guess I got carried away by the big, beautiful crystals and one of them turned out to be a 'kitchen sink/garden-whatchamacallit'. Nevertheless, we also drink strong coffee on this side, sometimes it helps, sometimes it doesn't, but as both of you have a knack for explaining difficult concepts in a comprehensible manner, I think I understand now (this is a long shot, but you don't maybe, just perhaps, speak Flemish/Dutch?? - would help a lot with my understanding - I wouldn't even need strong coffee 😃 ). Sad about the Cougar, though.

    Posted

  • ForestPreserve by ForestPreserve moderator

    Thanks all, and I see now that I managed to miss the related earlier discussions on the counting cars thread.

    Posted