CV suggesting some hybrid taxa, but hybrid taxon pages do not display "Computer vision model: included" badge

The CV is once again suggesting the hybrid Anas superciliosa × platyrhynchos for photos of mallards in North America, even though that is completely out of range, and last I heard hybrid taxa were supposed to be excluded from CV (has this changed?).
Here are some of the observations in question.

Also, the taxon page for the hybrid does NOT have the “Computer vision included” badge in the sidebar, which makes me wonder if it is unintentionally included.

(This is the badge I refer to - this screenshot taken from the Mallard taxon page):
Screenshot 2024-07-05 115529

Also, based on this post, by @thebeachcomber it is not the only hybrid taxon currently being suggested by the CV.

1 Like

Based on a quick scan, most (if not all) of these appear to be coming from Seek. The iNaturalist website doesn’t do it. (I haven’t tried the app.)

As life’s annoyances go, I guess this is minor, but it is annoying. (I don’t object to hybrids being suggested, but for sure this one specific hybrid should not be suggested to everyone who posts a picture of a bird that looks like a Mallard from anywhere in the world.)

1 Like

Seek does run on an older CV model that might include hybrid taxa so that would make some sense (and might not be a bug). The linked post above, though, shows a hybrid suggestion from the CV on the iNat website which uses a different CV model.

1 Like

Correct, Seek’s model will still have hybrids in it, so it wouldn’t be a bug if the ID was made in Seek.

Which specific observation is that?

1 Like

Interesting! It does appear that Seek accounts for all the recent CV Anas superciliosa × platyrhynchos suggestions.

However, the Garden Strawberry hybrid, Fragaria × ananassa seems to be being suggested by CV in all of the apps as well as the website:
https://www.inaturalist.org/observations/identify?reviewed=any&quality_grade=needs_id%2Cresearch%2Ccasual&created_d2=2024-07-05&created_d1=2023-02-01&taxon_id=55366&createdDateType=range&place_id=any

Using this observation as a test, the website suggested the hybrid as an ID when I tried:

It seems to be fairly accurate with the strawberry suggestions, but it does make me wonder how many other hybrids are still getting CV suggestions that maybe should not be.

The post was this one
https://forum.inaturalist.org/t/new-computer-vision-model-released/24729/47

to
https://www.inaturalist.org/observations/57190587

Thanks, will take a look.

Looks like there was a bit of an edit war and when that taxon was exported for training it was ranked as a species. So, not a bug as far as I can tell.

5 Likes