Screenshot from a bird identification app with no actual spectrogram

This topic has already had a couple of discussions. Here are some links: https://forum.inaturalist.org/t/screenshot-from-a-bird-voice-identification-app-as-a-photo/42712

https://forum.inaturalist.org/t/spectrograph-allowed/31621/2

The Merlin app can identify a lot of bird sounds to species. The observation is a screenshot of the app’s list of birds identified. So, it could be argued the birds were identified by the app. Were they identified correctly? I have used Merlin for over a year, know most bird songs well, and can say with confidence it is not always correct. Just this past weekend, as a bird was singing, it first indicated Blue-headed Vireo, then in mid-song indicated to Red-eyed Vireo. That’s a 50% accuracy for that observation. So assuming the app is correct is one problem with relying on the app’s report as evidence for an observation in iNat.

Another problem with this particular observation is that there are no actual spectrograms pictured. Yes, Merlin does include spectrograms, but you have to scroll through the recording to get each song’s spectrogram. This observation just used the summary screen, which shows the spectrogram at its start, so there is no way to even begin to guess if spectrogram is correct, if one can read them accurately enough to know what the spectrogram for each reported bird should look like. This particular user created an observation for each bird pictures in the screenshot, so there are multiple observations based on a screenshot with no sound, no photo, and no actual spectrogram.

It seems to me allowing observations like this provide very little in the way of evidence. Why not allow pictures of checklists from eBird then?

There’s another very recent discussion on bat sounds:
https://forum.inaturalist.org/t/acoustic-bat-records/64628/2

If there is no original recording you can downvote evidence of organism in the DQA.

3 Likes

I think a good example is Summer Tanager and Scarlet Tanager. Merlin will happily spit out both when it hears a Summer Tanager singing

1 Like

If the screenshot is from Merlin, then it will show copyrighted images.

Identifications on iNat are supposed to be the identifier’s own. It is fine to use Merlin as a tool in identification, but if there is no indication that the ID was subject to human judgement, it is not appropriate.

2 Likes

Oh, yeah, I’ve had that too. Merlin is helpful, but I see too many birders when I’m out birding that just run it and think that what it’s reporting is what’s actually present. It drives me crazy. lol.

The user posted the recording to the observations in question. So that’s kind of nullified the reason for my post originally. But now that I’ve listened to the recording and compared it to the list the user submitted based on what Merlin observed, Merlin missed a few birds. And I’ve listened to other recordings from the user that have lists attached and the lists also miss birds that are in the recording. I’m assuming the lists are based on what the user got from Merlin, meaning that Merlin missed birds that are pretty obvious in the recording.

I would ask the user not to upload Merlin screenshots. Staff have asked that these not be flagged for copyright, but also don’t want them posted, which puts curators in a bit of a “no man’s land” when dealing with them.

Users shouldn’t rely on Merlin for IDs definitely, but the reality is that many do.

If there’s only a screenshot of Merlin and no recording, I think it’s fair to downvote “Evidence of Presence”

1 Like

Edit: My comment is irrelevant to this post because I cannot read, apparently. Haha

I guess there is no better alternative, but I don’t really think this is ideal. Spectograms are evidence. The reason for them not being allowed on iNat is this:

Personally, I disagree with spectrograms being handled differently from other “grey area” photos (such as tracks), but if staff have said that that’s the way it should be, then I don’t want to argue about it.

Merlin is a useful tool. I’m using it about a year now and learned much about birdsongs and you can sometimes spot birds easier.

  • But, yes you have to be careful.

It showed me oriolus oriolus sometimes, it shows regularly “Singdrossel” a trudus but not blackbird.

I don’t collect them in my personal checklists out of inat as have been around.
I learned something about the song of “Singdrossel” and I can’t say I recognise this repeating song…

There are other birds, which it showed, which I haven’t count yet..

Once it for sure noticed a slamming car door of one of the neighbours for a certain water bird.
So, no… it is no real evidence..

But you can record with merlin and if it aren’t to many birds on it you can upload that audio to Inat.
I did that with an owl. :grinning_face:

And always be extra careful, if there is a bird like a starling around. - This year it seems to be a cool thing to imitate buteo buteo in my area.:rofl:

  • Actually, I think that is quite fascinating what and which birds they imitate and if that changes.

Do we have any data for such kind of things?

1 Like

This

is not relevant to the screenshots in the observation OP described and to which I was referring.

The OP noted:

and

I don’t ID spectrograms alone myself, but I also don’t downvote them for “Evidence of organism”. To my mind, someone agreeing with an ID on an observation with the evidence described above (a list of species Merlin thinks are present with no spectrograms visible) is just someone saying “I believe Merlin” and not IDing based on their own evaluation of the evidence of the species.

1 Like

You’re right, sorry. I skipped over the post and must have missed it