• Welcome to BirdForum, the internet's largest birding community with thousands of members from all over the world. The forums are dedicated to wild birds, birding, binoculars and equipment and all that goes with it.

    Please register for an account to take part in the discussions in the forum, post your pictures in the gallery and more.
Where premium quality meets exceptional value. ZEISS Conquest HDX.

Started using Merlin's Sound ID, holy smokes it's a game changer (1 Viewer)

Was this is Staffordshire?
Assuming your thumbs up means you were in Staffs I can only say it's highly unlikely you saw a Short-toed Treecreeper. Only a rare visitor to UK, with most, of about 40 records, from Kent.
More likely, as was suggested up thread, that you heard a Coal Tit and saw a Eurasian Treecreeper.
 
Assuming your thumbs up means you were in Staffs I can only say it's highly unlikely you saw a Short-toed Treecreeper. Only a rare visitor to UK, with most, of about 40 records, from Kent.
More likely, as was suggested up thread, that you heard a Coal Tit and saw a Eurasian Treecreeper.

The only thumbs up on the post is not from the individual who posted the Short-toed Treecreeper.

Occasionally, it is the interpretation of the technology that can be flawed. 😀

All the best

Paul
 

Attachments

  • Screenshot_20241002_221933_Chrome~2.jpg
    Screenshot_20241002_221933_Chrome~2.jpg
    171.6 KB · Views: 18
My mum (in her eighties) discovered Merlin a few months back and has thoroughly enjoyed learning which garden birds are responsible for the songs she's been hearing all her life. She was perfectly well aware though that Bittern was a highly unlikely visitor to her back garden! In the end she figured out that it was a distant cow.

Along with others here, I would caution against bold claims based on Merlin only ID. Always remember, extraordinary claims require extraordinary evidence.
 
When I used Merlin in Europe in spring, where I know bird sounds well, it overlooked about 50% of birds, and made a mistake 3-5% of times. In practice, in about 10 minutes it claimed one bird which was not calling. Merlin records audio, so I can be sure this was a mistake, not a bird it overlooked.

When I used Merlin in Mexico, every interesting sound I checked first with the real recording of the species. If it was identical and distinctive, I could start looking for the bird visually. Still, it sometimes made completely false claims (I remember some Greenlet was claimed out of the blue - no similar bird singing), and many interesting sounds it did not name.

Unfortunately, Merlin does not learn - if a bird or a sound is not in the algorithm, then Merlin will always overlook that species, until a new version of the algorithm comes up. Which is a bummer for birders, and makes some interesting errors if ornithologists use data gathered directly or indirectly from Merlin to e.g. count birds.
 
When I used Merlin in Europe in spring, where I know bird sounds well, it overlooked about 50% of birds, and made a mistake 3-5% of times. In practice, in about 10 minutes it claimed one bird which was not calling. Merlin records audio, so I can be sure this was a mistake, not a bird it overlooked.

When I used Merlin in Mexico, every interesting sound I checked first with the real recording of the species. If it was identical and distinctive, I could start looking for the bird visually. Still, it sometimes made completely false claims (I remember some Greenlet was claimed out of the blue - no similar bird singing), and many interesting sounds it did not name.

Unfortunately, Merlin does not learn - if a bird or a sound is not in the algorithm, then Merlin will always overlook that species, until a new version of the algorithm comes up. Which is a bummer for birders, and makes some interesting errors if ornithologists use data gathered directly or indirectly from Merlin to e.g. count birds.
I just wanted to mention a few fine points, for whatever they're worth:

- regarding the overlooking of birds, this often has more to do with the microphone than with the software. I have learned several interesting things about my phone intake while using Merlin - for example if I start a recording when there is traffic nearby, the volume of the sonograms seems "calibrated" to that sound level. Then if I start again with less background noise, suddenly it starts picking up all the little passerines around (provided I'm not stomping in dry leaves or whatnot!) I've also had "background birds" drowned out by louder calls such as Blue Jays or American Robins, and I needed to do the same trick once the loudmouths have had their say. Directionality also seems important - it makes a difference if I'm pointing the phone speaker directly at the birds, or having the phone face me, or pointing at an echoing hard surface like my car. I expect that some phones have a vastly better performance than my does and I'm guessing this differential could make a huge difference in someone's perspective of Merlin's utility.

- Along those lines I have tested Merlin on my phone against the Birdweather system (which is something of a data-driven system - I'd say more research oriented than bird list oriented; I encourage anyone to check them out). The birdweather was better at high-pitched chip notes (e.g. White-throated Sparrow) but totally whiffed on a conspicuous Red-winged Blackbird song which was only given once or twice. Merlin was the opposite on these. Both systems missed a rather obvious but distant Downy Woodpecker. When I brought it up to birdweather, they had calibration suggestions for noise intake- (e.g. gain) as well as software attributes (e.g. confidence and probability). Those are things that are unmanageable to change for the Merlin/phone setup, but this illustrates the complexity of an AI sound ID system, and I'm not sure that many birders have had the opportunity to appreciate those subtleties.

- Merlin does indeed "learn" but it is based on entries to the Macaulay Library and eBird data, which is why using it is a different experience in Cape May, New Jersey as compared to Equitorial Guinea. AI algorithms are by definition in a constant state of upgrade, so its not that we're waiting on versions of it like a traditional software release. That said, it sounds like your point is that the algorithm needs to be "seeded" with data that stems from human intelligence and that is absolutely true.
 
In my experience it does short toed vs common treecreeper well. Never had the other suggested out of range despite quite a few readings of both.

Iberian chiffchaff on the other hand is a regular mistake
I’ve had it ignoring a point blank range Ruppells Warbler, repeatedly flagging Melodious Warbler where there was none and not picking up an infuriatingly unfamiliar finch/ bunting like call a few feet above my head last week.

At the same time it’s very accurate with Collared Flycatcher , alerts me to Savi’s Warbler and Penduline Tit when I’m oblivious with spring chorus and can pick out White throated Kingfisher at a distance.
 
I've found myself increasingly using Merlin to help with call notes and detection. Merlin can be a great learning tool, or it could inhibit learning (by relying on it and never bothering to understand the identifications), it all depends on how one uses it. The important thing to never do is think that Merlin can "confirm" an identification. That's backwards. The observer needs to confirm Merlin's ID with their own eyes and ears, and if necessary with the help of recordings or field guides that can explain the ID.
 
- regarding the overlooking of birds, this often has more to do with the microphone than with the software.

Actually, overlooked birds usually show up clearly on the sonogram produced, so this is not the case of wrong microphone.

Agreed that Merlin currently generally stops working when there is a background noise - one would think that filtering out such things as footsteps are cars should be easy.

Merlin also picks different birds with different sensitivity. For example, for me it struggles with a singing Starling. I presume it is because Starlings produce very dissimilar sounds - both high whistles and low knocking sounds.

What Merlin shown me is that simple calls of songbirds are often at the same frequency as their songs. I was surprised why Merlin can pick so fast some short unremarkable contact calls, then I noticed that.
 
Ruppell's Warbler isn't in the list of supported species, so it is not going to identify this,
It’s on the list of birds in the Turkey pack. I presumed that meant it was supported? Unless it was recently added or I’ve misunderstood the list purpose?

I should add I’m actually a big fan of Merlin, as an educational tool and a way of picking individual birds out it’s very handy.
 
It’s on the list of birds in the Turkey pack. I presumed that meant it was supported? Unless it was recently added or I’ve misunderstood the list purpose?

I should add I’m actually a big fan of Merlin, as an educational tool and a way of picking individual birds out it’s very handy.

More information attached. In fact only a relatively small number of species in the packs are supported.

All the best

Paul
 

Attachments

  • Screenshot_20241003_212611_Samsung Internet~2.jpg
    Screenshot_20241003_212611_Samsung Internet~2.jpg
    137.8 KB · Views: 17
More information attached. In fact only a relatively small number of species in the packs are supported.

All the best

Paul
Thanks Paul and Mark. 👍🏼
I’m guilty of using the app without really understanding how it works. It looks like the bird name found in the search function is derived from Trakus but doesn’t mean it’s automatically supported yet.
 
- regarding the overlooking of birds, this often has more to do with the microphone than with the software. I have learned several interesting things about my phone intake while using Merlin - for example if I start a recording when there is traffic nearby, the volume of the sonograms seems "calibrated" to that sound level. Then if I start again with less background noise, suddenly it starts picking up all the little passerines around (provided I'm not stomping in dry leaves or whatnot!) I've also had "background birds" drowned out by louder calls such as Blue Jays or American Robins, and I needed to do the same trick once the loudmouths have had their say. Directionality also seems important - it makes a difference if I'm pointing the phone speaker directly at the birds, or having the phone face me, or pointing at an echoing hard surface like my car. I expect that some phones have a vastly better performance than my does and I'm guessing this differential could make a huge difference in someone's perspective of Merlin's utility.

This is very helpful. I believe this is why sometimes the app is picking up everything vs. only a few of the birds. I would sometimes start the sound ID while my car was still running and then set my phone on the handrest, then get out and slam the door and go for a little nature walk. Very interesting, I'm going to experiment with this the next time I'm out.
 
I just was on a presentation about using AI for other purposes. It seems that there come new algorithms based on deep learning. They can classify similar tasks with much fewer examples. I am curious if Merlin adopts them? Gathering 100 recordings for rare birds, which only few recordings were collected, seems far fetched.

Another thought - will there be a similar app coming from China or some East Asian country? Currently Merlin does not deal with Asian birds. And they have lots of birders and IT skills.
 
My mum's Merlin once identified me making a sarcastic-disinterested 'mmm' response to her boring fact about a car park sign, as an Eider Duck. Nice to know I can do a very convincing Eider impression!
It's very satisfying to deliberately trick Merlin into identifying your impressions. I've done it with Scops Owl before. Give it a go!

In a year of using BirdNET, I learned more about bird vocalisations than people I know who have been birding for 60 years. I think at present these apps are more useful for learning than for identification.

Ditto what people have said about quality of hardware. My experience of Merlin SoundID has been deteriorating and I'm positive this is because my crappy phone has been deteriorating.
 

Users who are viewing this thread

Back
Top