Flagger | Content Author | Content | Reason | Flag Created | Resolved by | Resolution |
---|---|---|---|---|---|---|
treegrow | Greater Bee Fly (Bombylius major) |
Lots of misidentification due to erroneous iNat autosuggestions |
Oct. 20, 2019 21:24:51 +0000 | Not Resolved |
Katja, you cannot imagine how many such misidentifications we have to deal with here in Africa. The computer vision is, for various reasons, entrained primarily on North American insects. Hence we are swamped with North American identifications of our African insects every single hour of every single day.
And I don't mean just the wrong species, no! We deal with wrong genera mostly, and once I witnessed an AI-ID for an order that does not occur in Africa (Raphidioptera).
I live with it, since I understand what's going on. The only remedy that I can think of is to patiently explain (and re-explain, etc.) the situation to each contributor upon correcting or neutralising their misidentification.
Good luck!
Riaan
Oh, I can imagine. I ID Diptera worldwide and see it every day. We also have a lot of problems within the US where western species pop up on the East Coast and vice versa. The current implementation of the algorithm has fewer egregious errors than the first one, but we still have a loooong way to go in insects.
Oh, I think I see what you mean: the "Other species commonly misidentified as this species" on the taxon page. That's not indicative of the problem here, because:
See comment about uncorrected misidentifications above.
I think this section only lists cases where an observation that was at some point identified as B. major now has a community ID for another SPECIES. So it does not count any of the observations where the new community ID is at the genus level or higher, which is probably the case for most challenged B. major misidentifications.
Hi - if this is a set of misidentifications that identifiers are finding it hard to keep on top of, you could add to the wiki here: https://forum.inaturalist.org/t/computer-vision-clean-up-wiki/7281
I'm going to "resolve" the flag on this taxon because the taxonomy itself does not need to be adjusted.
@lera This species has been listed in the computer vision clean-up wiki for several years, along with a link to this flag. It is my understanding that we leave these flags in place until the problem is resolved, which it is not in this case.
Taxon flags can be used to point out and discuss all kinds of issues: taxonomic problems, common names, issues with cover photos, conservation status, establishment means, etc.
Guidance about the computer vision flags is provided in the Computer vision clean-up - wiki: "Use flags or journal posts on iNaturalist, not this topic, to discuss specific taxa, identifications tips, and find others to help reidentify these observations. See the examples that are linked below."
We link to these flags all the time when correcting misidentifications. Open flags like this invite discussion and alert conscientious users that there is an issue. There is plenty of evidence that they help to improve identifier behavior. What is the point of closing flags without addressing the underlying issue?
Thanks @treegrow - I had a narrower conception of how flags should be used, as 'flag for curation' rather than also 'discuss id issues'. I also found support for your use case here - https://forum.inaturalist.org/t/big-ol-backlog-of-taxon-flags/4464/14
(So I shall leave such cases well alone!)
@treegrow I will try to help to solve some problems - I think that we at least can give it a try to allow the algorithm to get better ;-)
Thanks @szucsich ! Efforts to clean up misidentifications do pay off. There are several species where the CV has been retrained properly and that have remained fairly clean ever since. One example that comes to mind is Holorusia hespera. This used to be one of the species that was suggested for just about any large crane fly uploaded to iNat, and there used to be hundreds of misidentified specimens from all over the world. Nowadays we only get the occasional misidentification, and it's fairly easy to keep on top of things by checking every few weeks.
Hi all,
I've created a YouTube video detailing the identification of B. major in the US & Canada, and I am hosting a Zoom meeting this Sunday (part of a weekly program I've been doing) where we will go through observations. Hopefully, we will make some headway on the existing observations, and my video will be a helpful educational resource for new identifiers.
The video and accompanying information can be found here: https://www.inaturalist.org/posts/63131
-Zachary
Bombylius major is common in North America and Europe. However, there are many similar species in the subfamily Bombyliinae. In order to tell them apart, you generally need a clear view of the wing, so you can see the infuscated leading edge of the wing. In B. major, the dark band is sharply outlined against the clear part of the wing. It is fairly broad, but does not extend beyond the midline of the wing. In most similar species the dark band is narrower or broader or it fades gradually, and there is no sharp border between the dark & hyaline parts of the wing. A good picture of the head showing the shape of the antennae can also help. Currently there are many Bombyliinae in the inat B. major collection, but also quite a few Anthracinae. Unfortunately, these often go to research grade because many inexperienced identifiers seem to like to ID this species.