Oh no, sometimes I do find it very useful myself: with common species of distant places when I know the family or tribe. In those cases it works great; but probably it is not the main use CV is aimed to…
1 Like
This didn’t really occur to me but this is a great point to make. The CV was not very helpful in Singapore or Nepal, and it’s not very confident with a lot of stuff in Utah, either. In NY, it pretty much knows everything I’ve ever encountered and only has wacky suggestions if the pictures aren’t too hot.
5 Likes
Utah was much more difficult to Inat that.
1 Like
I use every possible way to get an accurate ID. You learn what the CV is good at the more you use it. You also learn about other identifiers knowledge as you add observations. I always research identifications for my observations, but sometimes there is no way to ID something without human help, and I have always found people on iNat who are more then willing to educate and help others on the site.
7 Likes
I think iNaturalist’s AI/CV is among the best I’ve seen out there, and I’m super grateful for its help getting me to a general ID for taxa I don’t know much about. I hope it gets better. There are taxa I don’t post or photograph as much because there are just so few identifiers, and AI can’t help. Whatever you might say about AI, it shows up and gives you an opinion, even if it can’t give you a specific ID, which humans just can’t always do.
For almost all my observations I type in the species. The AI can make some wacky guesses for birds, which is what I know best. And I’ve had awesome, hard-working people in some of the taxa I don’t know, with fewer IDers, correct an AI mistake after many months, but that’s rare. Dual shout-out to iNat AI AND all the human IDers out there who help people learn and make the data set better.
5 Likes
While I have mixed feelings on the iNat CV, I do appreciate it, and I also want to shout out eBird’s sound ID, which almost never leads me wrong and has helped me spot more than a few lifers.
1 Like
I really only ID Birds for context.
If the Bird is fully in frame and the photo is of a good quality (not a 10x digital zoom on an iPhone) I will usually Trust the CV 90% of the time. It’s usually right down to the Genus and species most of the time. However a lot of the times users do not upload the best quality photos and overly on the the CV to ID extremely grainy photos. I also wish there was a warning or indicator that a species has NOT been seen nearby when using the CV to identify; as I have seen a number of people use the CV to ID a bird and it recommends a species that is not native to that Country or even continent. I know it will say “Visually Similar / Seen Nearby” but I think it needs to conversely say in something like Red letters “Not known to be in this area” or “Not Native to this Location”
Overall, I usually rely on humans for rarer or more confusing species like Warblers or Gulls. If it’s a fairly common bird or species like a Crow or Squirrel I think CV can get it right most of the time.
1 Like
If my photos are really clear and the group has a lot of species that are in the computer model, like birds, butterflies, etc, I’m more inclined to use the computer suggested IDs.
Otherwise, for groups where there may not be as many species in the computer model, or if my photos are terrible, I use my own knowledge.
Sometimes AI can give a result which match with my idea, which save my time to type the name. I can say this is an example of combination. To me it would be the best way.
7 Likes
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.