Bird Observations With Only Merlin Screenshots

I have a question about how to handle bird observations where the only media uploaded is a screenshot of the Merlin app with a list of bird suggestions from the app. i.e., no audio file is uploaded, no Merlin export is attached- just a screen grab of the Merlin app suggesting that a particular bird is present.

  1. Is this considered a copyright violation, since it includes thumbnail photos of the birds suggested from the Merlin app that are presumably copyrighted?
  2. Is this considered “evidence of organism” at all by iNat’s standards?
  3. Some of the screengrabs also include a sonogram in the image, which might presumably be identifiable by someone who knows sonograms, but if the screengrab as a whole represents copyright infringement, should they be left up at all?

I don’t want to go off on a tangent about the pros and cons of Merlin as a tool, because I know that topic has been done to death, and I know properly uploading a file from Merlin is possible and encouraged (see: https://forum.inaturalist.org/t/importing-merlin-ids-to-inaturalist/37120) My question is specifically about observations where the only media uploaded is a screengrab from Merlin that says Merlin ID’d some sound as a particular bird, without the sound itself being uploaded. I won’t link to any of the ones I’ve seen, because that would be “calling out” users, but I’ve seen a lot of them recently, and I’m wondering what to do with them. Mark as “no evidence of organism”, flag for copyright infringement, leave them be if they have a sonogram in the picture…?

Thanks in advance!

2 Likes

I don’t see why spectrograms wouldn’t count as evidence, especially given many users interested in bats have been uploading them for years just fine.

2 Likes

If it’s just information from Merlin and no direct information about the bird, there is no evidence of the organism. It may also be a copyright violation.

If the sonogram displayed is from the bird observed in the wild (not a Merlin file for comparison) it’s OK. Maybe not optimal from my perspective, but certainly allowable.

Probably best to communicate with the observer about this. Find out about the sonogram. Point out that Merlin suggestions in themselves aren’t evidence of the bird. If violations of iNaturalist standards are happening and continue, contact the help desk.

9 Likes

I’ve uploaded spectrograms myself as well, I’m just curious if the inclusion of the Merlin thumbnail images in the image below the spectrum is a case of copyright violation or not. I don’t want to tell someone to change what they’re doing if it’s not considered a violation of anything, but I honestly don’t know whether it’s a violation or not.

1 Like

I’ve posed this question before and there were some differing opinions: https://forum.inaturalist.org/t/spotting-copyright-violations/3287/72
Staff never weighed in.

Without a scale, spectrograms should not count as evidence of an organism, and the Merlin app does not include a scale on it’s spectrograms. There have been numerous threads discussing why unstandardized spectrograms shouldn’t be uploaded and why they don’t count as evidence. So at the very least, I think it is absolutely appropriate to mark no for evidence of organism.

6 Likes

This kinda thing has been discussed before: https://forum.inaturalist.org/t/screenshot-from-a-bird-voice-identification-app-as-a-photo/42712

I don’t know the answer to the copyright question – while the images of birds in the screenshots are copyrighted, they may be a minor enough inclusion to be “de minimis” or “fair use”. Screenshots of non-free software are also technically non-free, but I think that might be a tough thing to enforce considering how many screenshots of gallery apps and photos of camera UI interfaces there are on the site.

When it comes to the “evidence of organism” question, I’d usually lean toward “no” if there is no audio. The screenshot of the Merlin screen may not even include the spectrogram of the bird in question (it shows a scrolling subset of the spectrogram as you play the audio), and the Merlin ID is not at all guaranteed to be accurate. If you can 100% verify that the bird in question is included in the spectrogram, it technically is evidence, and I would leave it marked as such, but it’s definitely not optimal (especially since the Merlin app doesn’t display a scale, although I’m pretty sure it’s the same 0-11khz scale as the Macaulay Library.) However, what I’ve gleaned from the thread I linked above is that while there is no rule against Merlin screenshots without audio, and that it’s essentially personal preference how you want to handle it, it’s at least strongly discouraged.

If audio is included, though, then you can simply just review the audio to see if the ID is correct. The only real harm I can see that having in that case (other than potentially copyright) is introducing junk data to the CV model.

2 Likes

When I encounter these observations I use a general copypasta which addresses:

  1. Sonograms are difficult to ID alone: attaching the audio file is easier for receiving identifications.
  2. A simple guide on exporting the audio file from Merlin: clicking the export/share icon (and describing the appearance in plain language) left of the play button so the file can be sent elsewhere. Offering to help with the process further if more guidance is needed.
  3. Informing them that the photos are copywritten and do not represent the organism observed, and advising them to crop to the sonogram or remove the image after adding the audio.

In the past, I have used this copypasta-style approach. I have asked curators to weigh in and they have chosen to remove full screenshot images based on copyright violations. I no longer report copyright violations in images aside from informing in a comment. Reporting makes the observation casual and a new user may be offput. This approach expresses more support for new users to upload the audio and get more positive experiences with the community.

5 Likes

Audio of the actual call should be submitted as evidence, IMO. That’s what can best be evaluated by others.

7 Likes

I agree that, without the actual audio, there isn’t evidence of the organism, and I use the DQA to downvote observations like this for that field. There’s a disconnect between presenting a visual evidence with both visual and audio information (spectrogram is audio in visual form). It also really isn’t possible to evaluate the evidence of a spectrogram effectively without any context and only unscaled visual information. So one legitimate approach could be to disagree to a higher taxon on the grounds that there isn’t sufficient evidence to ID to a species. This high taxon might be “Animals” since many animals could have made the noise, though, so I think downvoting for lack of evidence makes more sense.

There are also a lot of other issues:

A photo is uploaded that depicts an individual that wasn’t actually observed: this could throw off phenology and sex annotations. To users who aren’t familiar with Merlin, they might assume that the picture shown is indeed the individual observed and ID based on that picture. From this perspective, I feel the screenshot are a bit deceptive. Consider an observation that contained an image of a spectrogram and a stock image of a bird. We’d all agree that the bird image should be flagged for copyright and the observation would be casual grade. I don’t think the Merlin screenshot is effectively different.

Because of this, there’s also potential copyright issues, both for Merlin itself, and the photo shown as others have noted.

There are also issues inherent to any screenshots, especially multiple observations from Merlin that have repeated identical individual photos of the species from Merlin. This could lead to errors in the CV etc (though this is probably a more minor concern). Errors in the CV could also occur if Merlin screenshots themselves become prevalent as has been discussed in other spectrogram threads.

This also (to me) gets close to, though perhaps not over, the line for machine-generated content. There’s no indication that the observer has used any critical faculties to make their ID.

Also, it’s easy enough to upload the actual audio (about as easy as uploading a Merlin screenshot…)

Additionally, staff have asked users not to upload spectrograms.

In short, I think these type of observations violate multiple norms and best practices for how to use iNat, and think there are multiple legitimate ways to address via downvoting/IDing.

3 Likes

Without the audio nobody can positively ID anything solely on the Merlin app screenshot, and Merlin does get the IDs wrong itself often enough that it alone should not be considered evidence enough to be RG.

5 Likes

I came across a case during the recent CNC where an “observer” just uploaded a screen capture of the Merlin list of species ID in a song segment, but wasn’t aware that they could save, export, edit, and upload the actual recording. It’s a great app, getting better all the time, but like me with so many of the apps on my phone, there’s always a learning curve for novice users.
So a Merlin list of birds, without a sound file or spectrogram, in “no evidence” IMHO.

4 Likes

Thanks all for the replies! So it sounds like yes, these probably are copyright violations, and no, an unstandardized spectrogram should probably not count as evidence.

I’m curious for clarification about the use of standardized spectrograms now though- is posting them discouraged as well? I know my wife has posted some standardized spectrograms of single clear songs/calls that she’s obtained while doing her PhD work on birdsong, which can be easily identified by anyone familiar with spectrograms- no AI or Merlin or anything like that involved. Is this discouraged as well? Is the concern that it’ll mess with the CV modeling? Or just that the universe of people who can read spectrograms is much smaller than the world of people who can recognize audio calls?

1 Like

Yes, staff discourages (but does not forbid) posting spectrograms.

Standardized to what?

This is only true if the “anyone” involved happens to be familiar with the same scale. For example, I look at hundreds of audio files a day for work and can easily recognize my focal species in a spectrogram because I always use the same scale. But if you handed me a spectrogram on a different scale, unless it was very close to the one I use, I likely wouldn’t be able to identify my focal species.

I think there is a common misconception about what is involved in reading a spectrogram.

Because of this, no one can universally recognize a known species from just any spectrogram. The scale eBird uses is probably the one some people are most familiar with, but there is just too much variability. I’m not even sure if Merlin and eBird use the exact same scale even though they are both Cornell applications.

There is a feature request (Automatically add a spectrogram view to observations with sounds ) that if executed, would produce standardized spectrograms for iNat when audio is uploaded. But there are many logistic hurtles, so it probably won’t be implemented soon.

1 Like