Everything's coming up ...snowflakes

[GothHobbit]

Platform : website

Browser: Chrome Version 131.0.6778.267

URLs (aka web addresses) of any relevant observations or pages: from upload (https://www.inaturalist.org/observations/upload)

Screenshots of what you are seeing :

Description of problem (please provide a set of steps we can use to replicate the issue, and make as many as you need.): This may have been a transient issue, since I just retested with other photosets, and none of them are doing the same thing.

Step 1: upload photo(s) for new observation

Step 2: CV immediately adds the ID of Twisted Snowflake https://www.inaturalist.org/taxa/559988-Strumaria-spiralis

Step 3: ???

None of the photos have metadata or names that should bring up this CV suggestion. I’m also not certin how to replicate the bug; as I said, I just tried to replicate it with other photos, and CV didn’t auto-add any suggestions at all.

[tiwane]

Just making sure here: is the word “springs” in the metadata or file names of the photos at all? If so, then iNat is using that to fill in the ID, which it tries to do. “Springs” is one of the listed common names for that species.

[DianaStuder]

The same issue? (Why are there ducks on the taxon page for the flower)
https://www.inaturalist.org/observations/248580658

Thanks for resolving that ID.

[GothHobbit]

@tiwane, not the individual photo names. The folder name has Colorado Springs, but I didn’t think that the CV would pull from that.

[jnstuart]

That’s pretty twisted. And kind of flaky. But also interesting.

[wendyjegla]

I just asked my husband, a software developer, and he said it wouldn’t be typical for photo metadata to store path information (the path would include the folder), but it’s not impossible… and it is a suspicious coincidence.

I like these little mysteries, even if they’re technically problems.

[Vireya]

The Australian Wood Duck is at “Rankins Springs”. Co-incidence?

[wendyjegla]

The plot thickens…

[cigazze]

I’m guessing the word “Springs” is somewhere in the metadata of the image.

[jnstuart]

Could test this. Put a photo or two of some organism in a folder named Colorado Springs ( or just Springs) then upload to iNat to see what the CV does.

[jnstuart]

I tried this with the word Springs in the image filename (photo of a duck) and the CV suggested Twisted Snowflake. However the CV did not do this, and correctly IDed the image to species, when I put Springs in the folder name but not in the file name.

[GothHobbit]

@Vireya, @jnstuart; you could be right. The photos do have the location name, which is probably what’s triggering the CV auto-fill.

Something to keep in mind for my future work flow. Thanks!

[leafybye]

Why would the CV use metadata at all, what’s the rationale?

[tiwane]

CV doesn’t. iNat will read metadata and if keywords or the file name matches a taxon, iNat will add that as the intended ID. But no computer vision is involved.

Some folks like to add organism names in keywords or file names, and this speeds up the process for them.

[leafybye]

That could be helpful, I understand now.

[jnstuart]

In addition to this particular example, I could see this being a recurring issue if you name your photos for locations that correspond to organism names. Something to keep in mind.

[DianaStuder]

As the Wiki About says - Knysna is a town. So the observer labels their holiday snaps ‘Knysna’
But - also - on iNat - a Genus of monkey beetles - with one obs.

[tiwane]

We have a list of words that don’t generate ID suggestions, eg California and Argentina, which are both plant genera. If you have suggestions of words that should be on this list, please let us know