Better use of location in Computer Vision suggestions

I like the idea over-all, but it would have to be implemented with extreme care.

It should not exclude species based on location, as there is an ever increasing amount of invasive and non-native species spreading all through the world.

The ID suggestion should include non-local species, but perhaps with a geographical note along with it.

Something like: “This looks like XXYZ, which is generally found in eastern North America”


Has anything been done regarding this issue yet? Since the “seen nearby” feature already exists, wouldn’t it be an easy fix to restrict auto-suggestions to these ones only? If the users can tell that it’s something not on the list, they can still search it normally. The problem is that an overwhelming amount of users just pick the first thing on the list, which is often something from the wrong continent. As it stands, the system prioritizes species from high population centers (California, etc.) over anything else, regardless of the “seen nearby” function. The algorithm is clearly not good enough to detect an invasive species unless it is an extremely common species somewhere else anyway, so I wouldn’t worry about that being an issue. That is more of a rare occurrence and best left up to individual users.

Hi @saucegandhi, welcome to the iNat Forum. The staff recently responded here in this topic with this update - see above:


Thanks, I missed that one.

1 Like

The ‘Seen Nearby’ is also skewed bycultivated plants not being marked as casual. Then a domino effect from that first wrong one …