How to help CV recognize and correctly suggest Canna x hybrida?

General
1

If I’m understand this (https://help.inaturalist.org/en/support/solutions/articles/151000170368-which-taxa-are-included-in-the-computer-vision-suggestions-) right, this means that in order for Garden cannas Canna x hybrida, to be part of the CV, that means more people need to be agreeing that an observation is the garden canna, even if it’s a captive plant / casual observation?

I just went though over two thousand observations for Canna flaccida, and weeded out 1,969 observations that were actually Canna x hybrida. Turns out, many of the hybrid garden plants are called “golden canna” as a generic common name as well, which means people were misidentifying them as C.Flaccida by accident.

To help fix this, I changed the photos for both species to better illustrate the differences between them, and added “golden canna” to the list of common names for C.Hybrida. I also put the cover photo for C.Hybrida to be the most common form that gets misidentified for C.Flaccida, which is bright yellow with red speckles.

But just now I opened the uploader to make a test observation just to see how it would look if someone were uploading a photo of a Very Obvious C.Hybrida, and set the location for multiple areas where misidentified plants were posted. But none of them are suggesting C.Hybrida as an option, and aren’t even suggesting the genus Canna. When it does suggest any Canna species, it’s suggesting C.Flaccida, C.Glauca, and much more rarely, C.Indica.

Even though the photo does not actually match any of them, and the location is far, far outside where C.Flaccida and Glauca would be expected.

So, how do we fix this problem? If the CV is trained on captive observations, do I just need to find some other Canna lily enthusiasts and get them to help identify captive observations?

This observation here has a public domain photo of the seemingly most common form of C.Hybrida that is misidentified, if anyone wants to make their own test observation. You can also just use any of my photos of C.Hybrida.

7 Likes
2

The CV (almost?) never suggests hybrids - I think they’re deliberately excluded rather than just a lack of training.

7 Likes
3

And while there are no doubt good reasons for that, including iNat’s policy of preferring observation of wild organisms, this is one of those taxa for which the majority of observations are likely to be the hybrid.

3 Likes
4

Yes - it’s annoying for taxa mostly in cultivation as hybrids (Canna x hybrida, or Petunia x hybrida, or…) as well as naturalised/invasive hybrids like Crocosmia x crocosmiiflora (by far the most common in much of Europe!) or Reynoutria x bohemica.

3 Likes
5

I’m hoping they’ll make an exception for things like these, especially because I haven’t even gone throught the C.Glauca observations yet, and I already know a ton of them are also garden cannas misidentified…

3 Likes
6

There’s a feature request for this: https://forum.inaturalist.org/t/make-computer-vision-include-hybrid-taxa-on-an-opt-in-basis/60160/29

11 Likes
7

Added my vote! thanks for letting me know!

2 Likes
8

A lot of times I see garden flower taxon names with “hybrida” straight up deleted for not being real taxonomy. For example Hippeastrum hybridum was removed last week. So keep that in mind as you put effort into using Canna x hybrida as an ID.

3 Likes
9

Where I live, it is either Canna x hybrida or Canna indica only. Canna x hybrida is valid in POWO. Said to be ( C. glauca × C. indica × C. iridiflora ). Mostly seen in parks and home gardens. Canna indica is at wild places or beside canals. Big flower varieties will be in x hybrida. Earlier hybrids and wild species in America may be tricky to id.

3 Likes
10

The exclusion of hybrids can be problematic. For example, the vast majority of spruce in southern British Columbia are known to be natural white/Engelmann hybrids (Picea glauca x engelmannii), but that is not an option the CV model will ever suggest.

2 Likes
11

well that’s alarming. Especially because I finished going through both C.Flaccida and Glauca weeding out the obvious hybrids…and marked all the obvious hybrids that were not noted to be feral as captive.

12

these are not usually equivalent — the “hybrida” sorts of taxon entries being deleted are ones used for horticultural hybrids with complex ancestry. in the field of nomenclature (not in the colloquial sense of “naming” but the actual thing unto itself) these are not real nothotaxa, i.e. hybrid taxa with known ancestry, and are used as “trashcan names” that cover a wide range of actually different entities. Canna × hybrida on the other hand is a nothotaxon with standing — while it is a triple hybrid, it’s a known triple hybrid that is always the hybrid of those three parents. “hybrida taxa” usually are not used that way and therefore get deleted because they don’t refer specifically to a single entity.

6 Likes
13

Seems to me any name for a garden hybrid could become a “trashcan” if the parentage of cultivars isn’t well known/understood/documented. Are all cultivars of Canna really Canna × hybrida as defined, and if not then does every identifier using that name know how to tell which ones are or are not Canna × hybrida ?

14

it’s not relevant whether all observations of a nothotaxon are correctly identified, or we would see curators deleting any taxon entry with a large number of misidentifications. it’s that what I’m referring to as “hybrida” names are inherently taxonomically unidentified. again, the parentage has to be specifically defined in order to qualify as a taxon with standing, and in turn an iNaturalist entry has to reflect an actual taxon or clade with some standing. even an iNat “species complex” is not a taxon per se, but still has to be a monophyletic clade, per the guidelines. a “hybrida” name refers to a mess of hybrids that may not even have the same parentage, and therefore cannot be an actual taxon or nothotaxon, and in turn cannot have an iNaturalist taxon entry. “trashcan name”, just to reiterate, refers to the name (and by extension a taxon entry) and not what people are identifying as an entity. names of taxa, too, are misapplied all the time, and taxonomists sometimes have to step in to clarify the legitimate usage – but these kinds of “hybrida” names by definition do not actually apply to taxa, and therefore do not have a legitimate usage on iNaturalist in the first place.

1 Like
15

Oh yes of course misidentications aren’t connected to curration and I didn’t mean to argue with you on that point.

1 Like