What is the best way to get proper ID on those hard to properly ID species that are so similar EX: Anisota (oakworm moths).
I have been reaching out to top Identifiers of the specific species I am looking to ID and that has extremely helpful.(Thank you awesome people for your insist and knowledge)
Using the computer version and sifting through the genus or high taxonomy has been helpful as well. I’m looking for a source that can give me this is the key way to ID this species EX: it has and extra this or lacks this or whatever the key Identifying trait is. I think it would be awesome if there was a link to a data base that keys into the species and shows the Identifying trait of that species against it closely related counter part.
I signup a few days ago. still adjusting to the site figuring it out though. I like that they have links to books and works cited. I have already started a wish list of books.
Can someone explain how the AI software determines the ID from the picture. Does it look at just the first picture. Does it look at all the pictures (I have seen some observations that had different flowers uploaded on different pictures and the ID has been the of the first picture)? Is it important what order you upload the pictures? Is it important to have a finger in your picture so the AI software can determine the size. Should the picture containing a hand or finger be the first picture? Thank you in advance for some answers!
Welcome to the Forum!
The Computer Vision (CV) uses the first photo to make suggestions for IDs.
INaturalist is designed to handle one individual of a species per record. You can upload multiple pictures of one individual per record. So, if you have several flowers you want to post , make each one as a separate observation.
As @teellbee said, only first photo of the observation is analyzed by the CV model. All the model does is look at your photo and tell you that it looks like iNaturalist photos of certain taxa. It doesn’t analyze scale (eg fingers in the shot) or anything else, it just says “this photo looks like photos I have been told are of these taxa.” Also, by default, only visually similar species that have been seen nearby are shown as suggestions.
The only way your finger in the shot would help with getting an accurate Computer Vision ID is if the model was trained on many photos of that taxon with fingers in the shot. I wouldn’t worry about that. Take good, diagnostic photos. iNat is first and foremost a community and what’s best is to provide evidence for you and others to evaluate.
Finally, keep in mind that the computer vision suggestions are just suggestions so take them with a grain of salt. It should help you get started with more research if nothing else. And not everything can be identified to species from photos, which is OK.
One trick I use sometimes is checking the CV suggestions for all photos in the upload window before combining the pictures that go together into one observation. That tells you if there is some sort of consensus across all images, and sometimes is informative about which types of pictures work well and which are confusing for the CV. If you just want to check a picture without creating another observation, you can also using the CV demo page.
In the mobile app you can scroll between multiple uploaded photos and the app will offer suggestions on whichever photo you have currently showing. I like to scroll through all photos first if I’m not sure of ID, to see if the app suggests the same ID for every photo & angle. Maybe a little faster to do it that way than above suggestion if you don’t mind using the app instead
Thank you! That is a useful tool for when I’m just strolling in a garden and want to check captive/cultivated plants. I’ve bookmarked it.
If you find one of these, please add a disagreeing ID that is the lowest common denominator of all the organisms depicted (e.g., “flowering plants”) and then politely ask the observer to split the observation. These observations are more insidious than a simple misidentification of one species versus another species in the same genus.
PS. I wish there were some way to search for such observations (with some greater-than-random probability). I would then make it my mission in life to correct each one :-)
I bump multiple species back to Life. Until (if ever) they are split up to one species = one obs. A broad ID (flowering plants) tempts the next helpful identifier to ID as that’s a rose. And a daisy. And a lily.
You should but mark it “as good as can be” if id is changed to that taxon, so it goes to casual, that way others won’t be iding it anymore. (and of course add a comment about that, so you will be notified when other ids/comments are added, so you can withdraw the mark)
In other words, it identifies the same way a lot of humans do. Especially the ones who use photo-based field guides.
The definitive series of technical publications for many North American moth groups is the Moths of North America (MONA). Up until this past year, those have been expensive, hard copy publications to add to anyone’s library. Newer issues still are. However, the Wedge Entomological Research Foundation has now made all of the older issues available online as free PDF downloads. The volume covering the Saturniids, including Anisota, is MONA fascicle 20.2a. A link to the downloadable PDF can be found here:
or from any of the species’ pages on Moth Photographers’ Group, such as:
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.