Recently, I noticed that there’s an option to identify items using Google Lens. I’m using the Chrome browser — all you have to do is right-click the image and select “Search image with Google Lens.” I was wondering if anyone else uses this feature. Is it useful, especially in situations where iNaturalist can’t identify what’s in the picture?
Personally I’ve never found Google lens to be helpful in identification – it’s much less specialized than iNat’s own AI, for instance it doesn’t know anything about location. However I do find it useful for searching up a photo that I’m suspicious of being a copyright violation.
Sometimes if iNat CV suggests only one taxon for my own photo but I’m skeptical (e.g. its some sort of invert I know the CV struggles with/might have a lot of species below the CV threshold) I’ll do a google lens search to look for other things it could plausibly be (along with more traditional methods like checking close relatives of the species, the inat “similar species” tab, and looking at bugguide or whatever)
Edit: Also with inverts sometimes iNat gets completely stumped and offers clearly wrong IDs (often because the actual species is again below the observation count threshold for an ID suggestion) and here I will sometimes use google lens as a starting point (but trying my best to practice due diligence and research myself before submitting my ID)
I second jfmantis, only ever really useful for potentially copyright violating items and LLMs are really not reliable aside from extremely common and or conspicuous species, plus the cv is pretty decent along with the similar species tab.
I use it from time to time. One advantage is the ability to specify which part of the photo you are interested in. When there is a lot of extraneous stuff in the image, the iNat CV will give a lot of weird suggestions, whereas with Lens you can say that it is just this small part of the image that I am interested in, and have more hope of it finding something somewhere that looks similar.
That could be a good enhancement for iNat computer vision—to allow an identifier to select a cropped area of an image and have the computer vision engine return suggestions based on those pixels. I’d envisage this being presented via a “Crop” icon on the Suggestions tab in the Identify interface. When you click it you can select a portion of the image displayed to the left, and iNat will then return visually similar suggestions below that are based on the image portion you selected.
It would seem that the back-end functionality is already in place, as CV can produce suggestions for almost any size of image. But this would need a moderate amount of UI work.
When the CV completly fails because the invertebrate is too obscure, I will often search my image (On the upload page) in google lens. The google lens always gives an incorrect “It is X species” but then scrolling down, there have now been a handful of times where one of the image results is of a closely matching observation from inaturalist (Which has far too few observations to show up in CV). Its no guarentee to work, but it has produced a few results that I wasnt able to get otherwise.
I don’t use Google Lens for identifying on iNat, but do sometimes use it in the field for a quick preliminary ID. The right answer is often somewhere in the results, but not necessarily the first result. Think of it as someone handing you a field guide open to the right section - you still need enough skill to look through and see which entry (if any) matches the organism in front of you. It’s also somewhat biased toward suggesting more popular species first, so watch out for that if you do decide to use it. In general it’s much better with manmade objects than it is with nature.
So the short answer is: use iNaturalist to get help identifying the species. I’ve noticed that Google Lens sometimes pulls results from iNaturalist, but it doesn’t seem to be trained as extensively as iNat. When I have more free time, I’ll run some tests comparing iNat and Google Lens, and if I find anything interesting, I’ll share it here. Thank you for all your responses!
Every time I’ve tried with common wild species is far off. Also it seems to be very biased towards US species since it doesn’t consider the location. Outside US, useful for common garden plants only.