Opinion on the computer vision icon?

As many of you said, I use CV almost all the time, especially when I don’t feel like typing the species name or when I can’t remember the English name (because I’m from Argentina :argentina:) of it, but that doesn’t mean I don’t know what species it is. What’s more, if I ever use the CV and I’m not sure, I add that I’m not completely convinced that it is that particular species.


as someone new to iNat, i find VC to be really useful, especially since i often recognize a species but can’t always remember the name. i’ve also been taking a lot of time to read the summaries and compare different species using VC. it’s a very nice resource! and i totally get if anyone corrects me on my ID, but when im in doubt i pick a less specific identification anyway haha


*CV, not VC, my bad lol the dyslexia isn’t helping today


The CV icon doesn’t mean anything other than “you used the CV for this observation”. It isn’t a mark of shame. There’s of course a lot that can’t be accurately identified by the CV, but it’s a good starting point for a lot of organisms and a great time saver for common species regardless of your experience level, it’s not shameful to use.


I would have to go out of my way to avoid using CV - it pops up before I can start or finish typing, and usually, with what I ID it’s right anyway, so why wouldn’t I click it? I don’t think you can assume anything about someone’s knowledge just based on the CV usage.

Actually, in my experience identifying, just as often I’ve seen observations posted with an initial ID that hasn’t used the CV, and the ID given is incorrect. Then when I go to select a species, the CV suggestion pops up immediately with the correct species. (Usually happens with people wanting to believe the birds they’ve seen are rarer ones than they actually are, haha)

I mean, my usual iNat uploading flow is dumping all my pictures, clicking in the species box, and waiting for my CV options to click the one I know it is. I’m confident in my IDs. I’m not going to go out of my way to enter my data in a way that shaves off the CV symbol when the CV workflow is faster and more convenient. Also, it’s interesting to see what other species it suggests. People should be just as cautious when agreeing with a CV ID as they are with any ID - that is, you shouldn’t agree at all if you can’t identify it on your own, regardless of if the previous user used CV or not.


I’m perfectly fine with it.

Sometimes it’s correct and I know it is, so I accept it, other times I double check to see if the suggestion is likely correct, and if it is I’ll accept it (or go one taxonomic level up from the suggestion), other times I’ll start writing and it autocompletes to that, etc.

The idea that it is a “mark of shame”, or that someone “doesn’t know what they’re doing”, says far more about the person making tha assessment than it does about the person who made the observation.

In addition, when I see a mistake made in identification it’s really useful to see if it’s made from the CV or not as that helps me understand why the person may have made a mistake.

It’s worth remembering that no one will know all species, and that everyone will encounter species they are unfamiliar with, so even an expert may well find that the CV system is useful.

People should not be so eager to judge others.


In case it helps: you can type in a common name in whatever language you want and iNat will find it if it has been added to the database. This doesn’t affect what is displayed for other users – the ID will show up according to the preferences that they have set. So you don’t have to remember the English names (unless nobody has entered the common names that you are familiar with, but you can add these to iNat yourself if you choose).

However, I agree that for common species the CV is often faster than the time it takes me to recall the name (in any language) and start typing it, so I am not arguing that you shouldn’t use the CV if it is a more convenient way to get the name you are looking for.


But. Remember, it is YOUR name attached to the ID. CV is a tool you use, but it is your choice.


How different is using the CV from using a field guide reference (especially the glossy ones with photos)? My position is, not very much. For me these are both similar tools to achieve the same result: the identification of the organism.

Yes, sometimes using the CV can lead to an incorrect ID, but so does using a field guide.


I think that this definitely can be true for CV use, especially for many naturalists, but is not necessarily so. Using a field guide implies a base level of familiarity with the guide, needing to at least look through some options, and make a decision. It also implies owning or at least taking the trouble to access the field guide in the first place! This is something that requires some investment of time/effort.

Usage of the CV does not imply that effort or investment from the user. I think most folks on iNat do use the CV well, but there are some, especially new users, that may just pick the first CV choice without any consideration/comparison - this is something that likely wouldn’t happen with a field guide. As a tool, the CV requires much less investment and gives a quick, top-ranked answer which a field guide does not.

Relatedly, this is essentially what happens with Seek observations all the time - the ID is only what the CV suggests - there’s no manual input option, so I really like being able to see that an observation is from Seek.


Maybe this is off topic, but a few posters in this thread are testing the CV.
Do you have any idea how reliable the CV is for botanists?
More or less then 10% correct identified?

Generally, the apps were all better at identifying flowers than leaves, which the researchers say is due to their greater variety of shape and colour providing the AI with more clues. But this wasn’t always the case. The iNaturalist app was able to correctly identify just 3.6 per cent of flowers and 6.8 per cent of leaves. Plant Snap identified 35.7 per cent of flowers correctly and 17.1 per cent of leaves. The highest accuracy was achieved by Pl@ntNet at 88.2 per cent


M, very different! Keying out a thing, using clues and signs, or clicking on the first suggestion that “looks similar”?

Have no idea where they used the app, for my plant photos it’s around 80% when the right id is in the first 3 suggestions.


This is a misleading statement and is taken out of context from within the paper. In the results section it clearly states:

“…with 3.57% of flowers and 6.79% of leaves successfully identified to a species level at the initial suggestion. This is due to iNaturalist often preferentially providing a genus identification rather than a species identification” (bold text added by me for emphasis)

Also noting that all results in this paper were based on just 38 plant species from Ireland, which really couldn’t be considered a representative sample by any stretch of the imagination

I would have to carefully read the full paper, but from a brief skim it doesn’t seem like the highest quality piece of research (there are multiple significant typos already in the abstract alone; of course you shouldn’t condemn a paper solely on something like this, but gives an immediate bad impression of sloppy work here)



For example, plants that otherwise have an artic or alpine distribution can be found growing together or in close proximity to each other in the Burren in the West of Ireland. This makes the general flora a useful test case for plant ID apps. Photographs were taken at four sites: the Burren Nature Sanctuary, Kinvarra; Connemara National Park, Letterfrack and University of Galway campus all in Co. Galway and the National Botanic Gardens in Dublin. In total, 38 species across 15 families were included in the study (S1 Data)

It is from the newcientistwebsite, I found the list of species in the S1 excelsheet. I also saw the small amount of species, 38, compared with the tests of iNaturalist.

1 Like

And in fact there is already a comment on the paper noting this as well:

Disappointing that not only did the authors make this mistake, but none of the reviewers picked it up either, especially when it’s such an immediately obvious methodological flaw


Yes, I was referring to New Scientist being misleading, not you (sorry, I should have been clearer with my quote; hard on my phone)


Thank you for the comment, I missed it.


I find that since the last two updates the CV does a very good job on most common species in my area (Oregon, U.S.A.), including plants. Sometimes I’m impressed by its ability to name species like certain sterile grasses that many people would get wrong. However, when it’s wrong it’s often bizarrely wrong.


If the CV puts the correct ID down the list from very incorrect ones, I like to put in a manual ID. If I’m disagreeing with an ID on an image that I feel the CV would be very reliable with, I often go out of my way to have a CV identification.


Yeah, I get that. I only use CV in this way on my own observations—when I don’t know what something is, what choice do I have? I’d rather use CV to get a genus suggestion and increase my chances of a specialist seeing it, than leave it sitting as “fungi including lichens,” unidentified, forever.