[Ovah]

I'm very active on iNaturalist. The name suggestions made there by image recognition, at least for the genus I work with, is a complete and utter joke. >98% of name suggestions for that group are incorrect. As soon as a genus becomes diverse, it classifies everything as a single species instead of a genus. It does not have ability to recognize 'the unknown'. Most of my time on that website is wasted on clearing up gross incorrect name suggestions from AI that people accept without any checks of plausibility. Even with species that have >2000 confirmed observations, it still incorrectly suggests the name for obviously completely unrelated species. You didn't take the time to explain what you mean with 'hyperspectral' but I'm assuming it just introduces a new dataset where we start all over.

[thaumasiotes]

Seek by iNaturalist got a lot of very positive press here on HN for some reason, but it's crazy. I assume Seek is why iNaturalist classifies everything as a single species with no possibility of 'unknown' - as Seek views the world, the goal of classifying is to produce a species name, and accuracy isn't a concern.

I have a photo of an elephant seal that Seek informs me is actually a clouded monitor lizard.

I also have two photos of identical plants growing inches apart from each other that Seek informs me are unrelated.

[jfim]

Hyperspectral imaging is about collecting images in more than three bands of color (red, green, blue) as perceived by humans. Lots of things look very different in the IR and UV spectra, especially flowers and insects.

[taway789aaa6]

hah. tricked you! I rely on folks like you to help me ID the things that the app can obviously not ID correctly :) The best way to get the correct answer on the internet...

[s0rce]

I get quite a few genus level suggestions when it can't determine to the species level. I'm not an expert taxonomist so I'm sure there are still many overly specific incorrect assignments.