Make computer vision include hybrid taxa on an opt-in basis

[upupa-epops]

@Jarronevsbaru

It can be very difficult to work out the hybrids ancestry without genetic analysis as many cultivars have several species and cultivars in their ancestry. The highly popular cultivar Narcissus ‘Jetfire’ for example has 6 different species in it’s ancestry!

For some flower genera there used to be a generic “cultivated hybrids” taxon that situations like this could be identified as. I think those have been removed as invalid names now and messy plants should just be identified to genus. There was a suggestion for a different way to approach these here.

[silversea_starsong]

As others have said, nothotaxa =/= hybrids. Unfortunately iNaturalist does not have an ingrained system to differentiate if a taxon falls under the first or second case at this time. But nothotaxa absolutely should be part of computer vision, no question, as they operate like “real” species and it would be extremely beneficial.

[naturalist57011]

Anas platyrhynchos x Tadorna ferruginea?

[danielcahen]

My take is that, on balance, the benefits of including hybrid plant taxa in the model would far outweigh the costs. The current situation creates a significant burden on identifiers, as illustrated by the Kalanchoe × houghtonii example. Many common plant hybrids are misidentified as parent species, leading to a constant stream of corrections. This not only takes up valuable identifier time but also negatively impacts the accuracy of the model itself, as incorrect IDs are fed back into the training cycle.

Instead of adding complexity with an opt-in system, I believe a simpler and more effective solution is to just include these hybrid taxa in the regular training process and have them show up as suggestions like any other taxon. If a hybrid is visually distinct and common enough, it should be a valid suggestion.

[Jarronevsbaru]

That is a real shame, having a generic “cultivated hybrids” ID would make a lot of sense.

[Italopithecus]

I support your request.

[PdFuenteB]

I think this was discussed, and it was said that because GBIF and POWO do not support those taxa, it wasn’t possible. IMO, that makes a lot of sense for garden varieties and complex hybrids.

[rupertclayton]

@PdFuenteB

it was said that because GBIF and POWO do not support [generic “cultivated hybrid”] taxa, it wasn’t possible.

That might be overstating the limitation, I think. I believe this is a choice (maybe a good one), not a hard constraint. iNat can choose to make its own taxonomic choices, and even has a mechanism to explicitly state where iNat deviates from a taxonomic framework such as POWO.

I imagine that the decision not to have catch-all taxa for cultivated hybrids was based on an assumption that the effort of maintaining a large number of deviations was disproportionate to the value it would provide in observing nature.

[jasonhernandez74]

@upupa-epops

Which I guess makes sense since the [duck] hybrids, accounting for backcrosses, cover a spectrum of phenotypes between the two species. But it’s kind of misleading because hybrids are much less common than either pure species.

That may be true for ducks, but not universally for birds in general. For example, in the Puget Sound region, the Olympic Gull (Larus glaucescens x occidentalis) far outnumbers either parent species. The ID situation in this case is a similar mess to what @jf920 described for Kalanchoe, with hybrids usually being identified as Glaucous-winged Gull.

[nonbinary-naturalist]

You got my vote. I just went through close to three thousand observations for Canna flaccida (both casual and verifiable, since I plan to add plant phenology annotations, which include casual observtions in the data), and weeded out almost two thousand observations that were very clearly Canna x hybrida, with completely different colored flowers, leaves, and almost all of them far, far far outside any area where C.Flacidda would be likely to occur.

And when I’m done adding the plant phenology annotation, I’m going to have to go through C.Glauca as well. I already know a ton of them are actually C.Hybrida.

I even made my own thread about it.

[nonbinary-naturalist]

please this is absurd. Anyone posting an observation of a canna lily is given only the wild species options despite how blatant it is that the flower is a hybrid. I’ve had to subscribe to Canna flaccida and Canna glauca just to stop them from getting overrun again.

[jasonhernandez74]

@nonbinary-naturalist

I’ve had to subscribe to Canna flaccida and Canna glauca just to stop them from getting overrun again.

There seems to be a similar problem with Hibiscus x archeri (almost always initially IDed as Hibiscus rosa-chinensis).

[yerbasanta]

@wildskyflower

For example, in this T. x glauca observation, the CV reports 98.5% confidence (combined_score) for the observation to be T. angustifolia, but it shouldn’t. I surveyed the CV scores for maybe 20, and in many if not most of the T. x glauca observations I checked with the CV, the CV incorrectly had >~90% confidence for one of the two species. Reducing such unfounded confidence would inherently improve accuracy.

I know it’s not directly connected to the original request (which I agree with and voted for btw), but how do you see CV confidence scores? Is there a separate site for that or html code to access it?

[DianaStuder]

https://chromewebstore.google.com/detail/inaturalist-enhancement-s/hdnjehcihcpjphgbkagjobenejgldnah
Not the percentage as such - but it does show confidence levels.
One thousand users already.

Also Forum threads from when it was launched.

[jf920]

@tiwane Do you have an update on this? What’s the iNat-team’s opinion on the matter?

[tiwane]

I don’t have an update at the moment, I’m sorry. We’re pretty focused on the impending wider release of iNaturalist Next and then the City Nature Challenge.

[lmtaylor]

I’ve just added my vote for the modified version of the proposal where for plants inclusion would be opt-out not opt-in.