Is this what you mean?
From this topic: https://forum.inaturalist.org/t/add-a-flag-for-frequent-computer-vision-identification-errors/1905/3
3 Likes
hope I’m not completely off topic. Is this normal or I am just unable to use the interface properly?
1 Like
It is anothr topic, search for the right theme that was posted last weeks, such suggestions depends on what you was watching before opening this obs, better eplained at the theme.
1 Like
For “Out-of-taxon suggestions from visually-similar search with taxonomic restriction” see: https://forum.inaturalist.org/t/out-of-taxon-suggestions-from-visually-similar-search-with-taxonomic-restriction/7858/
“Filtering by both “Visually similar” and genus leads to odd results” also arose at: https://forum.inaturalist.org/t/filtering-by-both-visually-similar-and-genus-leads-to-odd-results/8295
2 Likes
That is exactly what I was asking for. I can’t believe it took me so long to notice your answer. Cool!
2 Likes
Will iNat deal with this in the future? It’s mostly brand-new users contributing to this- North America gets weekly reports of Gray Herons, Little Egrets, Brown Thornbills, Old World Buntings, Pacific Black Duck x Mallard, and more. It’s all well and good if users can manually fix the issue, but they aren’t the ones not checking ranges so it solves nothing.
it’s on the to-do list.
1 Like
It is a very difficult task to do programmatically. Doing it requires having detailed information at both a macro and micro level of geography. It requires tracking and storing that data in some way (currently iNat has a minimum of 5 different sources of data about what species are located where : its own observations, GBIF records, atlases, range maps and checklists and they are not consolidated or all linked together).
To use your example of Grey Heron is on the Canadian checklist (and the US one as well). It should be, the species has been seen in both nations, and a true checklist of the avifauna of both nations should include it. Should the species come up in Newfoundland or the Aleutians for example or still not suggested. Well, you can make it no, but that requires some way to mark that despite it being on the checklist and possibly even having records (I’m pretty sure without checking some of the Newfoundland records are documented as inat records) you still don’t want it suggested. Well, that then needs to be built and populated.
It is also really critical to properly deal with areas outside inat’s core user communities. In particular in Asia, Africa and to a lesser extent South America the distribution data is still poorly complete and a visual based suggestion tool may right now be the best tool.
2 Likes
I was running the identotron for here in the FSM and noticed that only Plumeria obtusa was listed. If the computer vision had checked the national list for Plumeria, only Plumeria obtusa might have been suggested. Yet both Plumeria rubra and Plumeria pudica are extant here. While I was able to update the national checklist to include these two species, I was reminded that internal sources such as the checklists might not always be complete. I have not yet updated checklists for Pohnpei island. In other countries propagating a species down to the many locations in a nation appears daunting at best. Thus eliminating possibilities by checking against the checklist could be problematic as the list may not be complete. As Chris noted above, expanding to include other sources would be difficult programmatically.
1 Like
You should never assume that a checklist is accurate or complete, but along with atlases (which bring their own set of issues) they remain the only way to document the presence of a species in a location where there are currently no iNat records.
Speaking as someone who added the distribution data for the over 50,000 species in the taxa group I curate into the relevant national checklists (I cant even imagine doing it for subnational levels), they are extremely time consuming to update. It is easier if you have a list that is geographically oriented, but doing for an individual species that needs to be added into multiple checklists - that is very slow work. Somewhere I have an active feature request to try and make it easier, but it got little traction, and I would have to find it.
edit here is the original requesr which generated a grand total of 1 votes
https://forum.inaturalist.org/t/improve-functionality-of-status-tab-on-taxon-page/3517
And what it was merged into
https://forum.inaturalist.org/t/improve-community-curation-tools-for-introduced-native-status/720/2
1 Like
The 2nd issue mentioned above has stopped being an issue in the most recent versions of the AI. It is now trained on higher taxa as well.
1 Like
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.