Search filter for observations that have identifications from Computer Vision suggestions

Platform(s): any

Description of need: The computer vision model is a useful tool but often is imperfect at identification. New users, especially with the City Nature Challenge, may rely/trust in the computer vision suggestions a lot more often. The results are many observations scattered throughout taxa that need additional attention quickly to reach a more accurate identification. Certain taxa are more curated than others, so observations that are suggested by computer vision to be a less-checked taxon may fall into obscurity. This can frustrate new users that find the app/site to be poor at identification (as indicated by a number of app store reviews that criticise the efficacy and speed of identification, both by the AI and the community). This may result in decreased user retention.

Feature request details: To expand on the already existing indication that an identification was made using computer vision, observations that have computer vision-aided identifications should be a filter option in search functionality. The icon exists already and could be added to the search filter UI. It can be separate from taxa filters so an identifier could look in their taxa of focus to ensure identifications are moving in the right direction.

I approved this request (although only for Identify, at the moment, as we’re currently not accepting requests for Explore) but I’m skeptical it will help you achieve your goals. There are a few reasons someone might use CV to add an ID - personally I almost always use it for my own observations, even if I know the species, because it’s easier than typing in a taxon name.

Also the funtionality as outlined by you here would search for observations where any of the IDs were added via CV, correct? That might not narrow things down enough to be usable, depending on the taxon.


Maybe iding for accounts that were created in a last week would be an easier way to find most of such ids?


My IDs also use whatever iNat CV tweak will get me to the name I want as quickly as possible. A list I can pick and click from, thank you.
rather … bie (for the common name)
than limp thru … ost mon (for Osteospermum moniliferum). And iNat is unforgiving of a typo of even one single letter.
Using a CV suggestion tells you nothing about whether that person has no idea what on earth it is - or they know exactly what it is and why it is that.