Since the CV runs only on the first photo, so long as the photo with the circles isn’t the first photo, it shouldn’t screw up the CV. But, this is a very valid point. Cropping helps a LOT and I do wish more people would do it.
just to clarify; while the CV suggestions only apply to the first photo of an observation when a user is requesting them, any of the photos in an observation could be used to train the CV
4 Likes
Ohhhhhhh. I didn’t know that. Today I learned something. Thanks for clarifying.
You’re essentially recording interaction, or relationship. To me, that’s what it’s all about - what are they doing, and why. And I have been known to take a series of photos, over minutes, or weeks, as something develops. But I think to most iNat’ers, it’s about cataloging individuals, almost without context.
In the past I’ve misused iNat by posting an image of a forest - maybe 300 trees of the same type. I was surprised that the iNat AI identified the species properly.
I have also commented like “I’m not interested in the … please help me identify the …”
2 Likes
Provide Cropped Photo
Would recommend only using this with new users and only if the organism is actually unidentifiable. Users are not required to provide cropped photos.
It’s helpful if you can crop the photo more closely to the subject. iNaturalist resizes images, so while we can zoom in to try to see it closer, the image does lose some resolution. Cropping usually makes it easier to get an identification too.
This is on iNat’s Frequently Used Responses page. I agree with earthknight that the purpose of iNat is to connect humans to nature. I do crop and use editing to highlight the species I want to ID, but those aids are directed at the humans who help ID my observations to make their job easier. By the way, thanks to everyone who identifies my observations!
3 Likes
Thanks it’s help me also !
1 Like
I agree, the “unsystematic manner” is key here. If certain species get circled more often than others, that may create a CV issue, but if it’s just random, it shouldn’t matter.
As someone who posts dissected moths a lot, I’ve run into the issue where the CV suggests the most-often-dissected moth species for all moth dissections, because it associates dissection images with those taxa. For example, if you click the CV suggestions for this one https://www.inaturalist.org/observations/331452675 you have Elachista, Eupithecia, Hahncappsia, Chionodes, and a Cochyline pop up. The gens in the image look nothing like any of those taxa, but those are the taxa that are most often dissected to identify by iNat users, so the image-matching assumes dissections are probably going to belong to one of those moths.
Besides the many plant/plant pathogen CV issues already mentioned, this absolutely happens with insects associated with a particular host or parasite. The CV suggests the fungus Hesperomyces harmoniae for all Asian Lady Beetle images, because Hesperomyces harmoniae is always in photos with Asian Lady Beetles. It always suggests the fly Istocheta aldrichi for Japanese Beetle images, because Istocheta aldrichi is usually in photos with Japanese Beetles.
There is zero doubt that if a “super-user” posted thousands of marked-up distant photos of woodpeckers in trees with the woodpecker circled, the CV would start calling all circles in trees woodpeckers. That’s just how image-matching works. But luckily, the “arrows and circles” crowd seems to be equal-opportunity for what taxa get circled at this point, so there’s no systematic application of these mark-ups, meaning that at present, they don’t seem to have messed the CV up at all. But regardless of how much importance you put on CV training, it’s undeniable that systematic circling of certain taxa at least has the potential to mess the CV up.
5 Likes
Certainly possible, or also possible that the CV learns that whatever is in the circle is what it should pay attention to.
Thanks so much for all these excellent answers and the interesting exchange of ideas. I looked more carefully at my observation and found that the feeding frenzy consisted mainly of robins, house finches, and one great-tailed grackle. I decided to post the picture and go with the suggested identification of the tree–an eastern redcedar–because it seemed like the most important organism in the picture. I had no idea that the tree’s millions of little red berries were such an coveted source of food for the birds–at least in early March. Thanks once again!
1 Like
You also get a bonus example of another thing the CV can struggle with: distant trees! Eastern redcedar has blue, not red berries. Actually, looking at your observation, seems like the CV is picking up on the telephone pole in front of the tree, which looks a bit like redcedar bark.
2 Likes
I agree with these suggestions and why. I also crop and use the mark-up tool to circle the species for each observation, then often also add in the NOTES something like “the circled species” then describe that species. Yes this is time consuming on your part, but the intention of iNat is clear concise quality observations so the viewers can accurately identify the species to move it to research grade material, which then might assist scientists and researchers based on the criteria they are seeking. For example, if I take a photo of a huge conglomeration of waterfowl on our local bay, I will duplicate the photo a number of times, crop some to highlight the species for that observation and circle the birds specific to that observation with a color that is in high contrast to the photo. If the photo is a number of light colored shore birds against a backdrop of highly reflective water/blue/silver, then I use a really irritating eye-catching color like bright pink or orange to circle those birds for ID. I usually include the original photo for perspective as an additional photo, but not the photo in first position for the observation.
2 Likes
Wow. I guess human beings aren’t obsolete after all!
2 Likes