Assuming there weren’t a stack of observations previously that were ID’ed as this species (especially given it doesn’t even have a placeholder image on the taxon page) → it got into the CV → they were all corrected (similar to Sarcophaga carnaria), how is this getting suggested?
Seek IDs this as Alpinia zerumbet, so I’m not sure if this is what happened in the above case, but looks like:
-Create an observation with Seek, which Seek has auto-IDed to species
-Share that observation to iNaturalist, but before saving it, manually change the species name to something else
-On iNaturalist your ID will still include the little computer vision tag even though you manually changed the species name
It would be interesting to hear from the user what their process was if they respond to queries.
I’ve come across a few like this I think. If the process is as @bouteloua mentions it might be - I’d still be curious to know what would cause them to manually select something so niche.
It does seem odd the computer would suggest a taxon with so few observations. Doesn’t seem it could have been trained on this. But mysterious are the ways of computers.
This post by @kueda describes a feature of the computer vision suggestions, that of “nearby injections”, whereby species with the same common ancestor as the top model results, if they have been seen nearby, can also get suggested. They don’t need to have enough total observations to have been included in the model. It’s a way of having relevant species suggested, even if there are as yet few observations. I suspect that is what happened in this case.
I expect Seek does not work this way. This talk does in my opinion not apply to Seek. And I expect that the species is not included in the Computer Vision model, but it is included in the final results.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.