Does iNat use other image databases, other external resources, for id suggestions?

A mycologist friend of mine collected a fungus, Sepedonium ampullosporum, on a property in Wake County, NC and [posted it to iNat (https://www.inaturalist.org/observations/62153710). To id it to species requires examination of microscopic features. He was very surprised to find that iNat’s id suggestion was, in fact, Sepedonium ampullosporum.

He notes two aspects that make iNaturalist’s correct id truly remarkable:

  • His record is the only one for S. ampullosporum in iNaturalist.
  • The virtual impossibility of field id by a person, even a professional mycologist.

Does iNat pull photos from elsewhere, he wonders? Even then, there are not many photos out there of this fungus. For example, there are none on Mushroom Observer nor on MyCoPortal. (Species Fungorum lists no synonyms for the binomial.)

Anyone have any insights? As for me, I wonder if the singularity is nigh :grinning:!

2 Likes

Hmm, his ID doesn’t have the little “magic tab” that should be there when someone has selected a computer vision suggestion.

image

It looks like he would have typed in the name, rather that selecting one of the automated image recognition suggestions. These are the top suggestions:

No, the model is just trained on other iNaturalist observations.

4 Likes

Welcome to the Forum, by the way :)

3 Likes

There is also the chance that the filename could have imported the ID correctly too, which would explain part of the confusion in thinking that iNaturalist did the ID. iNat pulls ID suggestions from filenames now, which leads to some weird IDs, but lots of helpful ones as well.

14 Likes

There is no taxon photo, so it must be pulling the filename from the photo?

3 Likes

Thanks, everyone!