In what manner do you use the computer vision (CV)?

I’m asking this out of general curiosity. Obviously, everybody knows you cannot rely on the CV as always being correct, but it’s still utilized and I think that’s great. Sometimes it causes problems when people use it but then don’t interact with other IDs afterwards or something like that. What motivated me to ask this though, is that I saw somebody comment recently that they thought only inexperienced users used it. This made me wonder what people’s reasonings are for when they do use it.

Personally, I use it when I have no idea what something is and the suggestions looks like it could be right. I know a lot of people don’t like that, but I usually make note of my uncertainty in the notes section if I choose to do this. Another time I use it is when I am uploading a lot at a time and I don’t feel like typing anymore. If the suggestion is what I was going to type, I just click it instead of typing it all in.

That persons comment got me thinking about this because as somebody who does use the CV sometimes, I didn’t realize that could reflect a possible lack of knowledge.

Anyways, the original question: how do you use CV or under what circumstances do you use it? Do you think there’s right and wrong ways to use it? I just want to know what people think about it. Thanks!


Only to laugh at the absurdity of some of its suggestions


For plants of the bioregion i live in (my area of expertise, i get paid to identify them) i often use it as a lazy workaround when I already know what the plant is, rather than typing in the plant, or to see if I think it gets it right. If I don’t know a plant, i will try the iNat algorithm. If it’s something obscure the suggestions will usually be obviously wrong based on the knowledge base i already have and i won’t use them, and it’s on to the dichotomous keys and such. However, an odd plant i find is often something that is much more common elsewhere in which case the algorithm saves me all that trouble.

If i have less expertise - say, plants but in a different area i am not familiar with, fungi, moths… i will use the algorithm, look at it to see if it seems right, maybe do some basic research. If it looks right but i really don’t know i will usually add the algorithm ID but note ‘not certain’ in the notes. If it seems sketchy i don’t use it and just ID to higher taxonomic level.

So yeah, i am user number 2179, meaning there aren’t very many more experienced iNat users than me other than Ken-Ichi, Scott, etc… and I use the algorithm.

People use iNat a lot of different ways. Some people focus on really ncie looking photos. Some people try to just get one of each species they see. Some people try to map as much as they can. You’re a mapper, I am too, some people find it annoying, but that’s how it goes on a big website like this. When i started here iNat was so small i basically personally reviewedall plant observations in New England and most of California in the US. Nowadays i am nowhere even close to there. Back then I felt like i could rely on most data points because if i thought they were wrong i would have downvoted them. Now I know most data hasn’t been reviewed at all, because the site is so big and growing so fast. Back then the data was more ‘reliable’ but it was pretty lonely on here… I like it better now. When i first started out I took a lot of blurry photos no one else could identify (many of which i just deleted or removed the photo for). Now i try to get diagnostic photos or just don’t bother with a photo if I can’t.

Anyhow i would say you are not doing anything wrong by using the algorithm as long as people aren’t then adding verifying IDs also using only the algorithm, that is pointless. I do think it is pretty sketchy for things like distinguishing spruce species from a blurry photo. Though perhaps all my old blurry spruce photos have helped train it. And, often it is surprisingly good, for plants. I know for some taxa it is pretty much useless.

The big thing is if you are gonna put a bunch of observations on iNat and rely on the algorithm you should stick around as they get identified to make sure to manage for ones that are wrong or need tweaking.

Also (edit to add this): I almost never use it for IDing someone else’s observations. Pretty much never if they used the algorithm to identify it too. What’s the point? I click reviewed and move on to the next thing.


I do mostly the same things. I’ll use it as a shortcut so I don’t have to type in names for plants I already know.

For insects where I very rarely know know what the organism is, I’ll often select the name chosen by the AI and then quickly withdraw it if anybody comes along with a disagreeing ID.

I also use it for identifying other observations when it’s something which I’m really curious about. If I do that I will do more research into the AI suggestion before actually submitting an ID based on the AI.

I use the AI with minimal critical review on my own observations, but only with great critical review on other people’s observations.


I use it as a quick way to get an identification I know is right - usually for common birds in my area, or one of the ubiquitous plants like Scotch Broom. Also sometimes to get a genus or family-level ID on something I’m not familiar with, like grasshoppers or small flowering plants. While identifying, I occasionally compare suggestions to find a likely genus-or-above ID (I do a lot of Unknowns, and sometimes I get tired of typing in “Plants” over and over again). But I never rely on it for taxa I know are difficult (like, say, moths) and I make sure to check all my notifications in case someone disagrees. My experience has been that its genus suggestions are often accurate for plants and insects, so I use them quite a bit when I find things I can’t identify beyond Order level.


As a guide to narrow down my search. Eg. if the CV suggestion suggests some spp. that are in a certain family I would look further in that family.


This is pretty spot on for me too, in terms of when you use it. It’s like it’s a shortcut and then helpful in times of complete confusion. I don’t doubt there’s other perfectly reasonable ways to use it, but I was wondering what they could be.

1 Like

I think the only time I ever use it for identifying other people’s stuff is if it’s at state of life or unknown, and then I’ll use it if it looks right or close to what the picture is of. I don’t usually go genus level because I tend to be clueless.


Not always. I use the CV like a field guide - that looks like This plant does the CV agree with me. It is another tool, but still up to us to decide if it is right, or not.


protip… “plants” IDs are a real pain in the butt since they include seaweeds and moss. Try to use a finer ID like fern, gymnosperm (pines), monocot (grasses, lilies, things with linear leaves) or dicot (for most everything else).

I only have my identify set to tracheophytes so I never have to see pictures of seaweeds or moss.

Or if that’s too complicated just use tracheophyte if it’s a land plant which isn’t a moss.


I do the same as @dallon and @charlie
Quicker entering of things I know
Curiosity to see if CV will be right (vs my own expertise)
and then getting in the ballpark for things I have no expertise on but usually noting something like “ID based on CV only” in description. I’ll also withdraw my AI ID if someone conflicts with it and if an inexperienced user agrees with the AI ID as well.


I always check the CV for my own observations, to see if it recognizes the species. I will usually go with its genus suggestion if it says that it’s sure unless I know that it is incorrect. I never rely on it unless it displays the message saying that “we’re pretty sure it’s this genus” or whatever.


For me the CV usually works as a shortcut for quickness of proccessing (that way I don’t need to type the name of something I already know for sure).
When I’m trying to identify something, sometimes I upload the photos and see what the CV suggests (before actually uploading the observation). When I have more than one photo of the same organism, I see what the CV suggests for each photo individually to check if it suggests the same thing for all of them. These CV suggestions help me guide and narrow my search during the identification process


Depends on what I’m ID-ing. For Lepidoptera it seems to generally do a good job as I don’t know many of what is generally well described. For birds I use it as ‘autocomplete’ as it generally gets it right. For parasitoid wasps it’s practically useless.

So it’s a bit ironic, it’s works best with species that are most well known and observed, but helps least for the more unknown creatures/plants that need ID help…


Curriculum vitae? What does this even mean?

I use it so I don’t have to type out the name a lot. Or sometimes if I’m familiar with the species but I can’t recall the name.

I use keys for plants a lot but sometimes I’m not even sure where to start. CV is very helpful to give me some suggestions of families I wouldn’t even thought to check (or had never heard of before). I have two books with keys but they only keys out to species or genera that are in the book. Same with keys online. Sometimes the list of suggestions are related and I can look at each one and figure out the highest taxonomic level that they are related (tribe, subfamily, etc) and put that as my ID.

With insects there are often things that look a lot alike that aren’t related. This is the most common time that I put down a suggestion that might be wrong and wait for someone to correct me. There are some people that really know their stuff, but if I put in a higher level ID, they may never see it.


Yes, I know “Plants” isn’t good, and I am improving my ability to add lower-level IDs on Unknown plants. I no longer use “Plants” - I’ve moved on to “Flowering Plants” and am working on learning which things I can safely ID as “Dicots”. I think I would have progressed quicker in this regard if it was more obvious what those terms actually mean, but I’m trying. Maybe I should look through everything IDed as “Kingdom Plantae” and nudge them down lower, to make up for all the times I used that ID as a beginning identifier :) Thanks for the tip anyway. I’m sure it will help someone else.


I do that too when I’m uploading observations. I check the CV suggestion before I combine the photos into one observation. Sometimes I will also use the confidence level of the CV to decided which photo to put as the first one.


I don’t think it reflects a lack of knowledge, but it came to my attention that some people see it that way, which I don’t like at all.


This is something I do too. I think that type of willingness is kind of essential if you’re going to be using the CV.