Did Merlin Bird ID really hear these in my garden? (1 Viewer)

[William Lewis]

Merlin's very useful. I use it quite regularly although know birdsong's reasonably well. I tend to use to mostly to narrow down an area particular birds are calling from but don't confirm an i.d unless it's something I know the call of very well or have seen.

Glad you've found the hobby!

Will

 

[Earnest lad]

...and that perhaps, in reality, weren't present.

Of course. It is all new to me. I am learning.

 

[Rotherbirder]

Artificial intelligence is just that - artificial! I personally wouldn't want the pleasure of the hard-learned skill of identifying vocalisations to be taken away from me by a 'robot'. I don't see the point of it unless your birding revolves solely around ticking off species on a list.

RB

 

[cheshirebirder]

Artificial intelligence is just that - artificial! I personally wouldn't want the pleasure of the hard-learned skill of identifying vocalisations to be taken away from me by a 'robot'. I don't see the point of it unless your birding revolves solely around ticking off species on a list.

RB

I started using it this spring purely because my hearing nowadays can’t pick up birds like grasshopper warbler, treecreeper etc. and I’ve found it useful. Agree that I’d prefer to use my own skills and it has thrown up some daft id’s ( Crissal thrasher in Lesvos ! ) but overall I’ve been impressed.

 

[Paul Chapman]

Artificial intelligence is just that - artificial! I personally wouldn't want the pleasure of the hard-learned skill of identifying vocalisations to be taken away from me by a 'robot'. I don't see the point of it unless your birding revolves solely around ticking off species on a list.

RB

An absolutely bizarre contribution in my view.

Sitting on a bench at Point Pelee during a rather busy minute, I ran Merlin and it enabled me to isolate a number of American wood warbler songs that I simply would not have picked up and I would probably otherwise have ignored. Nothing to do with ticking anything. Everything to do with learning the bird sounds around me.

It enhanced the experience. If anything, it also made the birding experience far harder work forcing me not to ignore any sounds.

It is difficult really to see much in this post other than a built in prejudice against something of which the poster is largely ignorant and a desire to take a bit of a swipe at other people only being interested in ticks.

I may be totally wrong and the poster may be more open-minded and less judgemental than appears from his post.

Indeed, he may have been able to identify all of the songs in the attached from Pelee without assistance unlike me. Many of the vocalisations appeared to depend on rhythm to my ear and some species had quite different variations of songs. The App allows you to click on the species identified and listen to the vocalisations from the library. It encourages you to question it and learn from the experience.

Edit - in contrast to the Pelee experience, the Moroccan recordings attached contained nothing unfamiliar. It was simply being run as a background check and enhancement to wandering along birding. But again, it encouraged me to listen to all the vocalisations.

Edit - the App also allows you to use it for your own recordings and to build up your own library. The poster may already have a means of doing this. I hope to use that functionality more in future. It will improve my own birding experiences. I made a start with the Ham Wall Great Reed Warbler:-

https://ebird.org/checklist/S136086271

Edit - a further point occurs. I have seen the App really enhance the experiences of casual or non-birders from nature walks by helping them on the sounds around them. That can only be a positive thing! Indeed, they may even become birders in the future.

I encourage the poster and others that think his post has validity to give the App a try in the right circumstances. A new world awaits that can enhance your birding.

The App will improve in its accuracy and the coverage will improve. Add you own recordings and you will be part of that process.

All the best

Paul

 

[LittleBitOfBreadNoCheese]

I started using it this spring purely because my hearing nowadays can’t pick up birds like grasshopper warbler, treecreeper etc. and I’ve found it useful. Agree that I’d prefer to use my own skills and it has thrown up some daft id’s ( Crissal thrasher in Lesvos ! ) but overall I’ve been impressed.

You can still use your own skills, just keep a list for what you saw/heard and a separate list for the automated IDs. But if I walk through a wood for two hours an AI can nail every audible call even if several are calling over each other, without bias. Up to the limits of its accuracy which are probably as good as anyone posting here. I can't throw away that kind of data. On top of that it can stay up all night without coffee or going to the toilet.

 

[Paul Chapman]

You can still use your own skills, just keep a list for what you saw/heard and a separate list for the automated IDs. But if I walk through a wood for two hours an AI can nail every audible call even if several are calling over each other, without bias. Up to the limits of its accuracy which are probably as good as anyone posting here. I can't throw away that kind of data. On top of that it can stay up all night without coffee or going to the toilet.

I would say that it does surprise me how often the App does not hear birds. I have not got great skills on identification by sound nor the best hearing but I am surprised by how much it misses.

It says that it supports Marsh Warbler but it certainly could not pick up the Yeovil bird on Thursday. I have often wondered whether an external directional microphone would help.

All the best

Paul

 

[cheshirebirder]

I would say that it does surprise me how often the App does not hear birds. I have not got great skills on identification by sound nor the best hearing but I am surprised by how much it misses.

It says that it supports Marsh Warbler but it certainly could not pick up the Yeovil bird on Thursday. I have often wondered whether an external directional microphone would help.

All the best

Paul

Yes , it’s odd that it sometimes fails to pick up birds that are perfectly audible to me.
Regarding some of the outlandish id’s , I wonder if the app is more governed by its American origins ? ( eg Eastern orphean warbler flagged as a thrasher )

 

[Paul Chapman]

Yes , it’s odd that it sometimes fails to pick up birds that are perfectly audible to me.
Regarding some of the outlandish id’s , I wonder if the app is more governed by its American origins ? ( eg Eastern orphean warbler flagged as a thrasher )

Eastern Orphean Warbler is not a supported bird so it presumably is looking for a best fit and is unaware of the confusion. Similarly, when it called Balearic Warbler as a Dartford Warbler for me. Balearic Warbler is also not supported.

I would upload my recording to eBird of the Balearic Warbler but I would need to edit the recording first as I can be heard quite audibly and clearly saying - "that's not a f###ing Dartford Warbler" to my birding companion when it came up with the identification. 😅

All the best

Paul

 

[Ruff-Leg]

Merlin can be a great tool for initial suggestions as to what may be vocalizing. The folks at eBird are pretty clear that it's just that - it's a suggestion, and it's up to the birder to confirm the ID. It's not Merlin's checklist - it's the birder's.

Merlin Sound ID best practices

Below are our recommended best practices when identifying birds with Sound ID in Merlin Bird ID. Following these guidelines will help to ensure your eBird observations appear in public outputs and are available for scientific research. 1. Th...

support.ebird.org

Merlin can come up short with a bird like a Northern Mockingbird (as it did for me recently with a Yellow-breasted Chat) because Merlin is analyzing the spectrogram (not the audio itself) in three second chunks. A mimic or bird with a lot of variability in its vocalization can fool Merlin. eBird developers acknowledge that.

As the poster of the Pelee experience points out, Merlin is excellent at analyzing all of the bird sounds, not matter where your attention is directed, or the state of your hearing.

More on what Merlin can and cannot do here:

Finally, check Merlin's settings page, and make sure you're using the appropriate bird pack fo your area, and equally important, not using those from elsewhere!

 

[Ruff-Leg]

Eastern Orphean Warbler is not a supported bird so it presumably is looking for a best fit and is unaware of the confusion. Similarly, when it called Balearic Warbler as a Dartford Warbler for me. Balearic Warbler is also not supported.

I would upload my recording to eBird of the Balearic Warbler but I would need to edit the recording first as I can be heard quite audibly and clearly saying - "that's not a f###ing Dartford Warbler" to my birding companion when it came up with the identification. 😅

All the best

Paul

Trim and normalize your recording with Ocenaudio audio or Audacity before upload - do be aware that neither has a 'fart filter!'

 

[01101001]

Artificial intelligence is just that - artificial! I personally wouldn't want the pleasure of the hard-learned skill of identifying vocalisations to be taken away from me by a 'robot'. I don't see the point of it unless your birding revolves solely around ticking off species on a list.

RB

My concern as well (that robots might end up being used for such purposes instead of humans), although until I get to the point of being able to identify a satisfactory number of vocalisations myself, I like to rely on it as a help. For example, not all species I want to be able to identify are abundant enough for me to build up experience the traditional way, and I find listening to bird sounds at home even less entertaining given that I can go outside and try to pick up sounds made by actual birds in real time thanks to an app.

EDIT: It doesn't have much of a problem with Marsh Warblers here, although it may take a while to catch up or, indeed, identify species mimicked by it instead (if the sound is faint).

 

[01101001]

I would upload my recording to eBird of the Balearic Warbler

||
V

cornell merlin id

I downloaded the cornell Merlin Id and found it inaccurate, I live in the midlands of england. On setting the programme in my garden it found a Kingfisher where there is no water and on a Pied Flycatcher which is scarce in my location. On other occasions it showed Long Eared Owl and water Rail...

www.birdforum.net

Finally, check Merlin's settings page, and make sure you're using the appropriate bird pack fo your area, and equally important, not using those from elsewhere!

As it was suggested by another person upthread, I also think Merlin's bird packs are only for the 'Explore' function, and that the Sound ID has a set number of identifiable species anyway (I stand to be corrected if the rather longish video above says otherwise).

EDIT: Sounds can be easily edited on the phone--you only need WavePad Free, which allows for unlimited trimming and a set number of normalisations (best done to -3dB from what I've read) per reinstallment. You can't see the sonogram, unfortunately, but Merlin shows that already.

 

[Butty]

I can't throw away that kind of data.

Try. Or at least flag it as 'Not reliable'.

 

[01101001]

Or write it in the comments: 'Merlin also identified ...', which is what I sometimes do when the app seems pretty convinced (and I want to have a full picture of birds present at a given location), but I can't nail the ID nonetheless.

 

[la_ravachol]

I would say that it does surprise me how often the App does not hear birds. I have not got great skills on identification by sound nor the best hearing but I am surprised by how much it misses.

It says that it supports Marsh Warbler but it certainly could not pick up the Yeovil bird on Thursday. I have often wondered whether an external directional microphone would help.

All the best

Paul

I used BirdNet before Merlin Sound ID came out and use them in conjunction with each other now as I find each has its own advantages and disadvantages. So, for example, BirdNet is far more sensitive than Merlin and usually picks up the things Merlin misses, but then it also picks up more background noise and has issues if there's a breeze. One tip for getting around this problem with Merlin is to record the sound with a different app and then import it into Merlin to analyse.

Regarding Marsh Warblers, I found Merlin couldn't ID Marsh Warbler when I tried this in Romania last year, but it was tricked by several of its mimicries.

 

[LittleBitOfBreadNoCheese]

I used BirdNet before Merlin Sound ID came out and use them in conjunction with each other now as I find each has its own advantages and disadvantages. So, for example, BirdNet is far more sensitive than Merlin and usually picks up the things Merlin misses, but then it also picks up more background noise and has issues if there's a breeze. One tip for getting around this problem with Merlin is to record the sound with a different app and then import it into Merlin to analyse.

Regarding Marsh Warblers, I found Merlin couldn't ID Marsh Warbler when I tried this in Romania last year, but it was tricked by several of its mimicries.

Do you mean it reports them as something else or nothing at all? I have a bN problem with Grasshopper Warblers, it seems to not think its a bird at all even if I stand right next to one. The species shows up on the birdNet stats page so it must work for some people at least.

 

[la_ravachol]

Do you mean it reports them as something else or nothing at all? I have a bN problem with Grasshopper Warblers, it seems to not think its a bird at all even if I stand right next to one. The species shows up on the birdNet stats page so it must work for some people at least.

It reports them as other species. I recall it identified an individual singing Marsh Warbler as Bee-eater and White Wagtail. I got the correct result with BirdNet, though it was sometimes coming out with Moustached Warbler and Paddyfield Warbler. Interestingly, I also got the correct result when I ran the BirdNet recordings through Merlin for analysis (also came out with Moustached Warbler too - Merlin really likes to call things this species). So it's not that it's incapable of identifying Marsh Warbler, but it's something about the way it records that is preventing it from doing so on some occasions. I find Merlin's recordings are not only less sensitive but there seems to be some kind of automatic filter to reduce background noise which I think can severely reduce the quality of the recording and often hinders getting any useful result.

 

[01101001]

Do you mean it reports them as something else or nothing at all? I have a bN problem with Grasshopper Warblers, it seems to not think its a bird at all even if I stand right next to one. The species shows up on the birdNet stats page so it must work for some people at least.

My experience with Grasshopper Warbler is that BirdNET treats every cricket like it's one. In this case, Merlin seems to score every time, no false hits (not that I know of).

 

[01101001]

Merlin has got much better recently. Are your versions up to date? Have you signed up for beta testing?