Computer vision clean-up - wiki

What is the desired outcome of this wiki? Are we asking other humans for back up, or is it implied staff will somehow edit the computer vision for these species?

5 Likes

So, what about species that are frequently misidentified in general, but have been aided by CV? The most obvious case I can think of is Taraxacum officinale. It wouldn’t surprised me if large numbers of them are actually T. erythrospermum or maybe even some other species I don’t know (in Texas, it wouldn’t surprise me if half are T. erythrospermum). I feel ill equipped to even attempt to correct these on anything but a very local level for several reasons: 1. Photos provided often (usually?) lack the required characteristics for species ID (free phyllaries and/or achenes). 2. I only know how to separate the two mentioned above when there are many more species listed for the US (and the world for that matter). 3. Species characteristics are difficult to use and/or interpret and/or are variable. 4. Even if the above three reasons weren’t problematic, the sheer abundance of observations seems unmanageable.

This could be contrasted with members of Euphorbia sect. Anisophyllum (another weedy group). While these are all frequently misidentified by CV and people alike, the numbers are low enough to manage (though, all the E. maculata observations and corrections are getting close at certain times of year) and I at least have a better understanding of them.

I guess my question is, since this isn’t really a CV problem, should Taraxacum be added here or is there another list that this might fit bit better on?

1 Like

Overwhelmed is a strong term. Not really feeling overwhelmed since the backlog has been dealt with, but more annoyed at still having to keep checking and correcting as they come through.

2 Likes

I just started the Mycetozoa section. Speaking of which, what is a preferred organization of the sections? It may have defaulted to “none” at the moment…

I don’t see the edit button.

Not 100% certain, but I think some Desmids are very often mis-identified.

Also, Stentor is never identified properly, definitely a problem.

The idea for the post came from seeing there were over 500 observations of a crabapple tree species (Malus fusca) outside of its known range. That seemed like wayyyy more errors than I was used to seeing for most species, and it reminded me of some previous similar cases that I’ve mentioned on the forum: Trombidium holocericeum, a red mite species that the internet had Decided was what all red mites were (and then CV too), which previously had hundreds of misidentified obs until I recruited a few people to help out, or Chrysoperla carnea, a similar case with green lacewing species, which helpfully has a species complex taxon now.

It’s most common with cryptic or just difficult to identify organisms; like @nathantaylor says, every dandelion is apparently Taraxacum officinale. Organisms that were misidentified in general and aided by CV seem to be some of the most extreme cases, so I would definitely include them here. People are generally more hesitant to refute an ID with a coarser rank than to refute it with a new species level ID. I don’t think this data quality task is as fun for most people.

So the purpose is to get help with clean-up. Maybe someone will browse the list and say “hey I know that species, I can help”, or maybe they want to pick one to learn something new.

I’m not expecting the staff to change anything at this time. They did do some supremely helpful tweaks to computer vision this year, since it will recommend different ranks like family or tribe when it’s not confident about species.

I’m trying not to break them into too many nodes, so I was just roughly grouping related organisms together and then alphabetizing them within the sections.

Looks like your “trust status” was updated, so you should see the edit button now.

2 Likes

Taraxacum has a problem that goes beyond identification. Taraxacum officinale was lectotypified with a specimen that isn’t what most people call Taraxacum officinale. The traditional usage has continued. Therefore, the name is totally confusing. Almost always misused (except occasionally by accident). I think we should just allow people to call the common weeds T. officinale and remind anyone who wants to use these in a research project that the name is a mess and its application is a mess and fixing it is beyond iNaturalist’s capabilities.

4 Likes

More about Taraxacum here: https://forum.inaturalist.org/t/overlooked-dandelion-diversity-in-bc-and-everywhere-in-north-america/3808

4 Likes

@else has suggested these to me in a message on iNat, but I haven’t checked to see if these qualify for “clean-up” per se:

Candy Cap - this name is used by the AI for any brown mushroom species; this is dangerous, as i see it applied to deadly poisonous species, whereas Candy Cap itself is a very nice edible. This name should definitely not be suggested at all.

other often misused names, but these do not concern me as much…
Russula cerolens
Russula sanguinea for every red-capped Russula (there are many red-capped Russula species that can’t be identified by a photo alon).
Macrolepiota clelandii - only occurs in New Zealand and Australia.
Leratiomyces erythrocephalus - a New Zealand species, but the name is used for anything red-headed on a white stem (young Russulas for instance)

I added Eurybia divaricata. I have this one constrained to its native range now (with one introduced occurrence outside that). But even within its native range the majority of observations are incorrect. Auto-ID seems to call a lot of broad-leaved Astereae E. divaricata

1 Like

I’ve been working on cleaning it up for Ontario, but the CV seems to think all hawthorns are Crataegus laevigata which is a rare introduction here.

I think the computer has learned Carex obnupta – more or less. The numbers probably aren’t large enough to add to the list above, but there is a problem with out-of-range observations (and not with the similar C. lyngbyei, which is also restricted to the Pacific Coast). The “Carex obnupta” observations from the east, Midwest, Sweden, China, and Palau do have leaves like Carex obnupta, but they can’t be that. I’ve been downgrading them to Carex (or Cyperaceae).

And what are the in-range Carex obnupta? Mostly Carex obnupta – people in the area who will post a Carex are often familiar with this common species. Also Carex lyngbyei, C. barbarae, C. kelloggii, C. puriflora, C. densa (odd choice), and C. mertensii, as well as Juncus and Phalaris arundinacea.

i just added ommatoiulus moreleti. i think that it would be decently easy to go through all the IDs and fix them but i dont have the knowledge to really do much more than the glaringly obvious

Computer vision gone mad! (Carex leptalea is a sedge, a plant, though a very skinny plant.)

https://www.inaturalist.org/observations/15971205

1 Like

So, if I’m not seeing the “edit” button, then I don’t have adequate user history/status to access editing, true?

Yes, that should be true… but looking into your account you already have the requisite trust level. Nothing in the very bottom right corner of the first post?


I would recommend adding Flavoparmelia caperata, Lepraria finkii, and Chrysothrix candelaris to the fungi list. These are all common species that seem to be applied to anything that looks vaguely similar.

1 Like

I would like to add Mercurialis perennis (a European plant) but I’m not sure about its potential distribution in Northern America. The Computer vision often suggest this plant for bushy green plants. Before adding it to the list, does anyone know whether it has been actually introduced to Northern America ? I could not find anything about it (my sources are quite limited).

USDA Plants does not list Mercurialis perennis.

I added a “Common Themes” section. Please add ideas there!