I’ve noticed this as well. What is the ‘correct’ way to ID a cultivar rose? I just leave them at genus, but I’ve wondered if there’s a better ID for them. Another one the AI loves to suggest is Rosa bracteata for every whitish colored rose.
Roses in general are a horrible misidentified mess. I only know a couple with any confidence, and it’s a full-time job just keeping up with the bad IDs on two or three species.
I see “Rosa x hybrida” used in various references as a catchall term for these complex hybrids. I see that “Rosa hybrida” (apparently using the convention here of leaving out the “x”, although I also note that iNat uses “Hemerocallis x hybrids” for daylily cultivars, a seeming inconsistency?) is available in the pick list. I hadn’t noticed this before but I suppose using it may be an alternative to just “Rosa” which is what I usually type, along with an explanation that it’s actually a complex hybrid cultivar.
I’ve noticed that iNat has a very strong bias towards suggesting “Primula pauciflora” for any Dodecatheon/Primula photo posted for this area and the surrounding large area (i.e. Alberta and the contiguous northern states), regardless of whether any identifying characteristics are visible in the observation record, which is often just a single photo of the whole plant taken from some distance away, probably unidentifiable beyond genus/section. This probably contributes to some false distribution data, when people simply select the suggestion, often I would guess without realizing there may be one or more other superficially similar species, and without awareness of the fine details that may need to be examined to distinguish them.
Another example is for Gaillardia, in which the iNat AI has a strong bias towards suggesting Gaillardia pinnatifida, which is way out of range for this area.
As noted previously, it seems the tendency for AI to “try” to be overly precise may be counterproductive in some cases. On the other hand, maybe the error will fix itself with time in the case of the Gaillardia, as more observations are made for this area and corrected as to species?
Since this topic is getting kinda long now, and Discourse just jumps to the latest reply once you’ve viewed a topic, just a housekeeping note of repeating the Read Me section that’s at the top.
I think the only way to get this fixed on the current AI regime is to get a very large corpus of observations of roses marked to “Rosa”, and then marked to “No” on the “Can the ID Be Improved” DQA. Then you have an RG Rosa. But they always also get marked planted. Then we’d need the AI to train on 2+ vote “RG” planted observations. I hope and assume that this will happen one day, so I go for “genus Rosa” + Planted + As Good As Can Be. But ! Your Idea of Rosa x hybrida is even better, so lets get the word out!
It already does train on these.
Yay ! Sweetness !
I mark double roses of all kinds as “Rosa” and add this comment: “This is a cultivated, hybrid rose. It has complicated ancestry and no scientific species name.”
Sometimes I’ll add, “If you want to know the name of this rose cultivar, it’s better to consult one of the websites or Facebook pages devoted to cultivated roses.”
just looked back on this and want to thank whoever added the earthworms. i think the problem goes well beyond computer vision. unfortunately i dont know the first thing about them so i cant be of any help. but i think in general, theres a LOT of wrong earthworm IDs but im helpless to figure it out
Earthworms can be ided only when they’re mature, so for any juvenile ided to species you can say it’s wrong. For correct Id you need close pictures of his foreparts (head shape, and generative part, can’t google its English name now) and count of the rings. (from head to the mentioned part and how many rings are covered by it).
this is my boilerplate text for worms:
sadly, immature worms (without clitellum “ring/saddle”) can’t be keyed even to genus, and there are 670 Lumbricidae species. If you look at worm keys, you’ll see that you need:
Right, when I found the key for our region I only could id this one set of my photos: https://www.inaturalist.org/observations/35749184 At the same time phone photos from 2 metres get rg of L. terrestris.
I’ve already added this to the wiki, but it could use some clarification.
Basically any orchid in the subtribe Oncidiinae (colloquially known as the “Oncidium” alliance) is ID’d by the computer vision as Oncidium sphacelatum. The overwhelming majority of observations that this happens to (especially the very many cultivated hybrids in this subtribe) are not Onc. sphacelatum but rather one of hundreds of other possible options.
The computer vision shouldn’t go below subtribe level for these, because the taxonomy anywhere below that is a mess.
Also everything in this group is some variation of “yellow and brown flowers” so the AI will just end up getting all the species and genera mixed up anyway.
Thanks, this is a species I have some familiarity with. I can’t spot all the mis-ID’s but I just went through now and found a large number of Symphyotrichum sp. mis-ID’ed as this, mostly S. cordifolia, some E. macrophylla, some Oclemena acuminata, some that I think are probalby Doellingeria umbellata, some wildly mis-ID’ed plants from outside the aster tribe, and some I was unsure of.
Got tired of IDing but I may go back to this later. I have been wanting to write ID guides on asters and looking through these would be good practice to prepare me to later do this.
I suggest that the automatic suggestions don’t show Lumbricus terrestris anymore for the reasons you just explained.
I added Pholcus phalangoides, for many years only a few species were widely recognised for a big part of the world, now the “AI” suggest this species for almost anything that resembles Pholcus, and people agree, and other people agree with first. Someone with free time and hard nerves should recheck all RG under this taxon and it shoudn’t appear that easily for taxa were close examination of both dorsal and ventral parts (in female) are needed.
Add it.
Gonna add Entodon seductrix, AI for some reason loves this moss and suggests it for anything, even Atrichum (that is hard for me, they’re really, really hard to mix up, they’re totally different). From lmost 3k of observations it’s possible that real Entodon is in less than 1/3.
I cleaned Salix discolor the Old World, but these pop up every day.
I just finished cleaning up Succinea putris from the entire Americas with the help of a few other people