Started using Merlin's Sound ID, holy smokes it's a game changer (4 Viewers)

[NoobKevin]

I'm very new to birding. I recently got a job working outside, and checking out the wildlife goes hand in hand in with what I do. Why not make the most of it. Anyway, I tried the Merlin's sound ID for the first time today. I couldn't believe how well it worked. It's amazing actually. This brought up a few questions for the birding community. The app ended up ID'ing a rare bird for the area. In fact, per ebird, in the "hotspot" I was at, this bird has never been identified there before. It wasn't even on my downloaded pack, but it still identified it. How often is the sound ID wrong?

Also, considering it was identifying birds left and right with only half of them visible, and sound is an acceptable form of observation, what is the general consensus on using the "Sound ID" to grow someone's life list. In reality, as most people probably feel the same way, I'd rather see them and possibly take pictures but, if we're going by the rules, all we have to do is hear them. It almost feels like I'm skipping over some steps.

Also, I guess I didn't realize how important sound identification was. Birds are hard to see. They are easier to hear though, duh! What % of birder's life list are made up of sound only identifications? It's got to be up there.

Thanks,
Noob

 

[temmie]

I'm very new to birding. I recently got a job working outside, and checking out the wildlife goes hand in hand in with what I do. Why not make the most of it. Anyway, I tried the Merlin's sound ID for the first time today. I couldn't believe how well it worked. It's amazing actually.

it is! That is, in the designated areas. It will hopefully improve but it's still far from covering all species in many parts of the world, but that will be a matter of time.

This brought up a few questions for the birding community. The app ended up ID'ing a rare bird for the area. In fact, per ebird, in the "hotspot" I was at, this bird has never been identified there before. It wasn't even on my downloaded pack, but it still identified it. How often is the sound ID wrong?

Often. But let's say the commoner the bird or the more distinctive the sound, the more reliable it is. The problem arises with small call / alarm sounds that are often very similar between (sometimes very different-looking / different family) birds.

Also, considering it was identifying birds left and right with only half of them visible, and sound is an acceptable form of observation, what is the general consensus on using the "Sound ID" to grow someone's life list. In reality, as most people probably feel the same way, I'd rather see them and possibly take pictures but, if we're going by the rules, all we have to do is hear them. It almost feels like I'm skipping over some steps.

You're skipping some steps, yes

Also, I guess I didn't realize how important sound identification was.

Loads of people, birding for decades, never even get into sound ID. So besides skipping some steps, you got at least one essential step 'right': sounds matter. A lot.

Birds are hard to see. They are easier to hear though, duh! What % of birder's life list are made up of sound only identifications? It's got to be up there.

Your list, your rules. Most (big) listers will only count birds seen though, but even that line is a bit messy, e.g. I count a bird I cannot ID on sight, but have heard and next seen (so the ID is in the combination of hearing and seeing but sometimes ID is hard just by view, sound will support that).
So your list, your rules.

Thanks,
Noob

 

[Farnboro John]

If Merlin tells you it found a bird not previously recorded at a site then it's probably wrong.

I make no comment on people choosing to tick on sound but ticking local rarities on unreliable software advice without visual confirmation is right out.

John

 

[Brad Hacker]

I use bird-sound ID codes--including Merlin--a lot. They can be completely wrong in surprising ways: ID'ing birds that are not present, ID'ing birds that are present by mis-IDing the sounds of other birds, IDing motorcyles or dogs as birds, etc. Knowing what birds are likely to be present, listening to the recording yourself, and seeing the suspect birds are all required to make meaningful use of such code.

 

[Owene]

One of things that people don’t seem to mention much in the constant debate about Merlin and apps is that it can really help train your own listening skills. I find written down descriptions of calls very hard to translate into actual sounds but having Merlin on in the background and associating your actual listening to the species names has brought my aural birding on much more quickly.

But yeah generally if it comes up with something too exciting it’s wrong

 

[Paul Chapman]

One of things that people don’t seem to mention much in the constant debate about Merlin and apps is that it can really help train your own listening skills. I find written down descriptions of calls very hard to translate into actual sounds but having Merlin on in the background and associating your actual listening to the species names has brought my aural birding on much more quickly.

But yeah generally if it comes up with something too exciting it’s wrong

Absolutely. It is a device that is brilliant for learning calls. Sitting on a bench at Point Pelee & seeing various birds come up & then learning to listen for their songs was a brilliant exercise.

All of the attached were correct in slightly over a minute & I would not have learnt them as quickly otherwise. Roll on better coverage.

All the best

Paul

 

[A2GG]

I recently loaded Merlin on my iPhone and used it for the second time today. I find it very useful already.

 

[Brad Hacker]

Other resources include

BirdNET Sound ID – The easiest way to identify birds by sound.

birdnet.cornell.edu

Chirpity - Identify Birds by Sound

Experience the magic of bird identification through sound with Chirpity! This desktop application harnesses cutting-edge Machine Learning to identify bird calls in audio recordings. Tailored for enthusiasts and researchers, choose between BirdNET or our native model, perfect for Nocmig -...

chirpity.mattkirkland.co.uk

BirdWeather - a living library of bird vocalizations.

Join a growing community of citizen scientists with PUC, our AI powered bioacoustics platform.

www.birdweather.com

 

[NoobKevin]

it is! That is, in the designated areas. It will hopefully improve but it's still far from covering all species in many parts of the world, but that will be a matter of time.

Often. But let's say the commoner the bird or the more distinctive the sound, the more reliable it is. The problem arises with small call / alarm sounds that are often very similar between (sometimes very different-looking / different family) birds.

You're skipping some steps, yes

Loads of people, birding for decades, never even get into sound ID. So besides skipping some steps, you got at least one essential step 'right': sounds matter. A lot.

Your list, your rules. Most (big) listers will only count birds seen though, but even that line is a bit messy, e.g. I count a bird I cannot ID on sight, but have heard and next seen (so the ID is in the combination of hearing and seeing but sometimes ID is hard just by view, sound will support that).
So your list, your rules.

I think I've started to see some misidentifications with the app. I can see why people just skip over the sound part of the identification. It can be very difficult and, for me, I wouldn't trust my ears. So, that being said, I will definitely need a visual on any sounds (whether I'm using the app or not) before I will be confident enough on a positive ID. Thanks for the reply.

 

[NoobKevin]

Absolutely. It is a device that is brilliant for learning calls. Sitting on a bench at Point Pelee & seeing various birds come up & then learning to listen for their songs was a brilliant exercise.

All of the attached were correct in slightly over a minute & I would not have learnt them as quickly otherwise. Roll on better coverage.

All the best

Paul

For the new birder, I'll see a bird. May or may not be able to get a picture. Then it's gone. If I can remember its characteristics and/or the picture is clear enough, the next step is going through a 300 page bird book. Then, when you think you might have the right ID.... you start second guessing and going through the book again or looking it up online. At least with this app, it can point out the birds in the area, then if you do see them, you have a list of a few to choose from and can make a more confident ID.

This app is a little strange though. Yesterday, it was working great (other than some possible misidentifications) but today it wasn't pulling as many birds up, and it was taking longer. I had a scrub jay fly right in front of me making some noise, the app picked up the sound but it didn't spit out a bird. It's like it just ignored it. However, if it works only part of the time, that's good enough for me. I'm sure it will only get better.

 

[scruffybirdman]

I hear lots more birds than I see and I use Merlin to ID most of the time. That said, I use it primarily to ID common birds that I expect to see in the area anyway. When I get something odd, which is not that uncommon, I take what precautions I can to verify the ID.

If I am not familiar with the bird, I will stop the recording and start another just to see Merlin will make the same ID twice. If it does, great, that it is a bonus. If I only get it on the first recording I will see if it identified the same bird more than once. Either way, though, I will go on to step two and verify the vocalization by ear and or spectrogram analysis.

Then I take it all together and decide if I can make a good enough ID... for me... because I set the rules. I may get a query from a reviewer, and I can always change my mind later, but if I put it down then I have vetted the "sighting" to my standards and tend to stick with it unless I prove myself wrong-- which has happened more than once.

 

[Helsinki Birder]

It's awful.
Both the sound and to a lesser degree, the photo ID functions, feeding into ebird have utterly polluted things with false information when it comes to Ireland .

Occasionally things even breach onto the main news services, with nonsense put out that should simply have been ignored.

A Field Guide To Birders - 2nd Edition

Long ago, when I had the Cork-Dudeing Blog active, I wrote a field guide to the various types of birders out there. I felt it was time to r...

helhathnobirdies.blogspot.com

Owen

 

[temmie]

It's awful.
Both the sound and to a lesser degree, the photo ID functions, feeding into ebird have utterly polluted things with false information when it comes to Ireland .

Occasionally things even breach onto the main news services, with nonsense put out that should simply have been ignored.

A Field Guide To Birders - 2nd Edition

Long ago, when I had the Cork-Dudeing Blog active, I wrote a field guide to the various types of birders out there. I felt it was time to r...

helhathnobirdies.blogspot.com

Owen

It's a two-edged sword, thanks to photo ID apps we've had occasional nature lovers posting pics of genuine vagrants (0.1% of the time) vs. 99.9% of mostly common birds mis-ID'd as rarities. More importantly, many species groups are now ID'd much more often (correctly) and this adds to the citizen science data collection. One of the best and (for me) most astonishing examples is mothing. I don't know about moths but have ID'd about 200 in the garden for the past 2 years, all thanks to my moth equipment and photo ID (obsidentify app). I hope to double that figure in the coming years, and there are many out there just like me.

so...
Is it better to keep it relatively 'clean' but miss out on some vagrants?
Is it better to have few data on species groups that aren't looked at often, but are looked at now thanks to photo ID?

 

[Ruff-Leg]

For the new birder, I'll see a bird. May or may not be able to get a picture. Then it's gone. If I can remember its characteristics and/or the picture is clear enough, the next step is going through a 300 page bird book. Then, when you think you might have the right ID.... you start second guessing and going through the book again or looking it up online. At least with this app, it can point out the birds in the area, then if you do see them, you have a list of a few to choose from and can make a more confident ID.

This app is a little strange though. Yesterday, it was working great (other than some possible misidentifications) but today it wasn't pulling as many birds up, and it was taking longer. I had a scrub jay fly right in front of me making some noise, the app picked up the sound but it didn't spit out a bird. It's like it just ignored it. However, if it works only part of the time, that's good enough for me. I'm sure it will only get better.

Merlin is only as good as the recordings that are in the Macaulay Library at Cornell. Not all songs and calls of all species are in there. Cornell relies on a group of staff and Volunteers to curate and annotate the recordings to make them ready for machine learning. The process is done in Merlin Vision. Maybe haters should step up and contribute to the effort with good recordings and some curation work?

Once the model is trained for a species - they need a minimum of 100 high technical quality records for model development - and deployed to the app, you must remember that the app is analyzing the spectrogram in 3 second chunks. If an easily recognized bird pops up and makes a strange noise, Merlin isn't going to magically ID it. And, it's important to have the microphone on your phone pointed in the right direction, stop talking and moving around, and try to isolate the phone from ambient noise as best you can.

There are only 1382 birds in Sound ID, with a focus on North America. At present. It's a work in progress.

Cornell is adamant that ID's should not be reported based on the apps suggestion - that's all it is. A suggestion. And one that is getting better.

No, I don't work for Cornell...

 

[kids]

Personally , I use it as part of the process of ID' ing , just as you would any other part of the overall "feel " of a bird. The other day , by a massive Yew tree, I heard a call I was sure was one I didn't recognize . Merlin gave short toed treecreeper . I spent 5 minutes confirming by waiting and watching so a new species for me . But it constantly gives Black Redstart when in areas that aren't remotely likely to be correct .
so for me , it is part of whether I feel confident of a correct id but not accurate enough to be the whole of it

 

[cafe birder]

Where was the short-toed tree creeper please ? That is which country ? Yew trees are good places for Coal Tits which call very like short-toed tree creepers. The latter is a very rare visitor to the UK

 

[Mysticete]

Merlin is great...I use it all the time. I never use it as the only record of the bird for my lists, but it can be good for just "trawling" the local birdsongs to pick out interesting birds that you might want to spend time doing a more focused search for, or confirming the ID of a empidonax or other difficult visual identification. And its great for learning calls...somehow I have a much better time remembering a call if I can match it to something actively calling that is being identified, then just reading a text description of the call or listening to a tape.

 

[jurek]

Cornell is adamant that ID's should not be reported based on the apps suggestion - that's all it is. A suggestion. And one that is getting better.

Amen - Swiss bird database ornitho.ch recently published a warning for birders not to put reports based only on the ID app suggestions. Such artifacts (fake records) can potentially really pollute the bird records - especially that novice observers outnumber the experienced ones, and for birds which are locally rare or extinct it can easily happen that almost all records inserted are mistakes of a poor software.

 

[Kirk Roth]

I think I've started to see some misidentifications with the app. I can see why people just skip over the sound part of the identification. It can be very difficult and, for me, I wouldn't trust my ears. So, that being said, I will definitely need a visual on any sounds (whether I'm using the app or not) before I will be confident enough on a positive ID. Thanks for the reply.

I think that above anything else, it is important to keep in mind that Merlin is a tool, just like your binoculars, your field guide, and your research. There are many newer or less experienced birders who fall into the trap of treating Merlin as magic - that it will identify the birds for you. Yes, it can provide suggestions, but it up to our human brains to follow up and/or confirm. Yes, it is a powerful tool, especially in well-birded areas, but any tool can be misused. I am an eBird reviewer - there is a special category in review queues that is basically for reports with no more evidence than "Merlin said so." That is how common of a problem this is.

Rare birds are decidedly NOT one of Merlin's strengths. Aside from the obvious (e.g the chances of a common bird making an odd squeak, or even the sound of footsteps etc. that the AI library mis-recognizes), the algorithm will also "weigh" recognition based on where it believes you are. As an example, I was in south Texas using Merlin in an area known to have a vagrant Crimson-collared Grosbeak. It picked up the bird's call... but repeatedly insisted that it was a Northern Harrier. If I were "using" Merlin exclusively for ID - well there is a rare bird that would have gotten missed.

The best use for Merlin is for the relatively common birds - that is where it shines because of the bulk of input to feed the AI. If a bird or a location is unfamiliar to the program, it will be just as confused as you or I with the unfamiliarity. Here are some examples of my own usage where I find Merlin really helpful:

- Skill-building. One of my favorite uses of Merlin is when I'm hearing that bird that I know I've heard before - the recognition is there... but what is it? The old way was to sort through your field guide apps (or cds... or cassettes... I'm not old or anything...) but now Merlin can provide me that reminder so much more quickly. It really helps with those difficult chip notes. But then I can focus on the unfamiliar call so that I can really incorporate it - I find that this works so much better in real time than later when I'm set with my other resources. You can absolutely use Merlin to become a better birder.

- Detection. As aforementioned - I'm getting older and a lot of those high-pitched songs get by me. But also our human brains are designed to focus on particularly familiar or loud songs and "filter" out the background - our phone speakers don't do that. As such, my ability to detect the little peeps of quieter sparrows, chippy warblers among feeding flocks, Brown Creepers and so on has very much increased when I use Merlin. If it picks up a surprise, then I listen for it and lo and behold it is there - I just needed to be paying attention.

- Travel. When I'm traveling to someplace unfamiliar, Merlin can be a goldmine for getting up to speed on the local songs and chip notes, which very much beats out the old methods of studying your brains out. You still need to study and you still need to verify, but Merlin can be an absolute shortcut when, for example, you need to know that the chip note you've been hearing all day are Olive Sparrows - if I want to see one I should be looking on the ground... or if I want to find something else I should ignore that chip for now. It can give you an idea of whether a bird is more common than it may appear if Merlin is picking it up all the time - maybe they just hide a lot. It can give you an idea of what to expect and anticipate; I feel like Merlin can jump-start your familiarity with a new location because I haven't had the luxury of time to really know the place - especially if I limit myself to only a visual understanding of what is present.

- Yes, ID. But use with caution. You can go back and re-listen to the call and check the sonogram against other sources. Even better, if Merlin tells you there is a rare bird then you can track it down and get more of a verification - at least it will tell you what to focus on. Personally, I've never had Merlin correctly alert me to anything more rare than an early migrant, but it is always our responsibility to verify anything unusual or unexpected. That is just as true with Merlin as it would be a fleeting glimpse out of the corner of our eye, or a short view seeded with "expectation" of a rarity. If I don't verify the result, I don't trust it.

I'm hoping something I've said above is helpful. But the bottom line is that Merlin is a tool - it can be wielded skillfully or incorrectly, and as with any tool practice and familiarity increase its usefulness.

 

[wheatearlp]

Personally , I use it as part of the process of ID' ing , just as you would any other part of the overall "feel " of a bird. The other day , by a massive Yew tree, I heard a call I was sure was one I didn't recognize . Merlin gave short toed treecreeper . I spent 5 minutes confirming by waiting and watching so a new species for me . But it constantly gives Black Redstart when in areas that aren't remotely likely to be correct .
so for me , it is part of whether I feel confident of a correct id but not accurate enough to be the whole of it

Was this is Staffordshire?