Thumbnail image Sriram Reddy | Macaulay Library

[Alli Smith] All right. Hello, everybody and welcome to tonight’s webinar hosted by Bird Count, India and the Cornell Lab of Ornithology. My name is Alli Smith and I’ll be speaking today from the Cornell Lab in Ithaca, New York and I will be the host of tonight’s webinar.

Over the next hour, a bird expert from Bird Count India and all of you in the audience watching will be competing against the Merlin app in a Bird ID trivia game. And we’re also celebrating the Great Backyard Bird Count, which begins next week.

The Great Backyard Bird Count is a four day long event that anybody anywhere in the world can participate in. And all you need to do is watch birds for at least 15 minutes, identify what you find and submit your sightings to Merlin or eBird. To learn more, you can visit birdcount.org and we’ll share a little bit more information later in this webinar as well.

Before we begin, I have a few notes for our audiences watching on YouTube. This is an interactive game and we’ll be showing polls throughout the webinar. You can join the fund by entering your guests for each question into the polls that will appear in the chat on YouTube.

One other note about tonight, we have a game show packed with trivia so we unfortunately won’t have time to answer questions from the audience. However, we do have a team of helpers behind the scenes in the chat so if you need technical support or have quick questions about the Great Backyard Bird Count, please do send your message in the chat and we’ll do our best to get to those.

If you have a question about Merlin or how to use the app, we also have a wonderful Help Center with questions– with answers to common questions and you can also submit a message there to us on the website. We’ll put a link to the Help Center in the chat as well just in case anybody needs it. And that’s it for notes. So let’s get started with trivia.

Welcome to Merlin Bird ID trivia. Our game is very simple tonight. We’ll have 15 questions where I will show a photo of a bird or play a recording of a bird song and whoever identifies the most birds correctly wins.

With us today from Bird Count India is Ashwin Viswanathan. And Ashwin is a mechanical engineer turned ecologist and a longtime bird watcher. In his early research, he studied rainforest ecosystems and later became increasingly fascinated by the power of birds as a means to draw people towards nature and the potential of citizen science to answer ecological questions.

He now works with Bird Count India, a group that encourages bird watchers to systematically monitor birds and helps generate knowledge from the data that emerges. Ashwin will be competing against Merlin tonight and will also be sharing some tips on how to identify birds. Welcome to the game show. How are you feeling about taking on Merlin tonight?

[Ashwin Viswanathan] Thanks, Alli. I’m quite nervous, to be honest. Let’s see how this goes. But I assume it’s going to be a nice journey through, maybe what we are going to see through, GBBC and some familiar words to us.

[Alli Smith] Awesome. I think you’re going to do great. So tonight, Ashwin will be competing against the Merlin Bird ID app, which was developed here in New York by the Cornell Lab of Ornithology. And Merlin uses machine learning technology to identify birds in photos and sound recordings. Also with us today is Sam, who is a machine learning engineer with the Merlin app and he’ll be sharing more about how Merlin works behind the scenes throughout the trivia game. Sam, how do you think Merlin will do tonight?

[Samuel Heinrich] I’m really excited to see how Marilyn does on these ID challenges myself. Marilyn’s photo and sound ID algorithms are really heavily tested to ensure a high level of accuracy but they can have some limitations and make mistakes. So it will be fun to see how they kind of stack up against an expert like Ashwin.

[Alli Smith] Awesome. I can’t wait to see how everybody does. We’re almost ready to begin but first, I need to share the rules. So I will share my screen here and we will get started. All right.

All of the birds in tonight’s quiz will be species that can be found in India. We’ll have 15 questions. We’ll have 12 photo based questions and then three sound questions at the end. And for each question, I’ll show a photo of a bird or play a recording of a bird song.

Merlin and Ashwin will have 30 seconds to decide on an answer and all of you in the audience on YouTube will have 30 seconds to choose your answer from a poll that will pop up in the chat. After those 30 seconds are up, everyone will share their answer and whoever gets it correct will get a point. Whoever has the most points at the end will win.

There are also so many exciting species to learn about. So for each question, we’ll also be sharing a link in the chat that brings you to the Birds of the World page for the species if you’re interested in learning more about any of the birds tonight. Birds of the World is an incredible online resource with detailed scientific species accounts for every bird in the world and it’s available for free for anybody watching in India thanks to our partnership with Bird Count India.

So let’s begin. The first half of the game is based on photo ID. So identifying a bird visually is a really important skill that Ashwin here uses every time he goes out birding. When he finds a bird, he’s probably looking for the size and the shape of the bird, along with colors or patterns, maybe the length of the birds bill or other visual field marks that he can use to identify it. Merlin is a little bit different. Sam, could you briefly explain how Merlin identifies birds and photos.

[Samuel Heinrich] Yeah. So Merlin’s ability to identify birds in photos starts with the Macaulay Library, which is the Cornell labs state of the art digital archive of Natural History photographs, audio, and videos. From this archive, volunteers drew boxes around birds and hundreds of thousands of photographs from the Macaulay Library and identified which birds were in those boxes.

Then researchers used these box images to create a machine learning model called a convolutional neural network that can identify birds in photographs. Convolutional neural networks are able to extract important features such as size, shape, and color from an image and they can recognize patterns in those features that correspond to a particular species. This type of model is also extremely useful for bird identification because it can recognize two types of features, coarse grained and fine grained.

So coarse grained features are larger scale features, I think where the white is on this bird on the right side of the image. And fine grained features are really detailed ones, like the shape of a stripe on the side of a bird for instance. So Merlin can use both of these features to differentiate different species.

[Alli Smith] Awesome. That sounds really similar to how I might identify a bird actually. Like using my overall initial impression of size or really bold color and then combining that with the more fine details to get an exact identification. That’s really cool. So let’s jump in with our first question.

Here’s our first bird. This is a species that you could find throughout the year across most of India. And there are a few other species that do look pretty similar but this one can be found in maybe more damp and more forested habitats than similar looking species.

So take a look at the photo and we’ll turn on the poll in the chat so you can vote for your answer. We’ll give everyone about 30 seconds or so to vote. And I’ll keep an eye on the chat in YouTube to see what everyone’s thinking.

All right, we have– the chat is open in YouTube. I see lots of votes coming in. Excellent. All right. And we’ll say time is up for voting. We’ll close the poll and we’ll see what the audience has to say. I see a lot of other type answers in the chat too. All right. It looks like almost the entire audience voted for Jungle Babbler. Ashwin, what do you think about this species here?

[Ashwin Viswanathan] I would agree that it’s a Jungle Babbler. And from these options, there are two very similar looking Babblers, Yellow-eyed Babbler and Jungle Babbler. And that’s often a confusing pair for us in India. But I think in this case, the distribution map really, really tells a story because Yellow-billed Babbler is not found in– Jungle Babbler is not found in some of the flatter arid parts of South India.

And if you can go back to the map, you’ll see that there are certain areas where Jungle Babbler is absent whereas Yellow-billed Babbler is found there in the South. That’s a good clue. Also, Yellow-billed Babbler is not distributed in the northern parts of the country. So that’s another important clue and helps us identify it as Jungle Babbler.

And I think as human, we at the moment have a slightly better grip of our probabilities. Maybe of where a species might be found or what’s likely to be in different places. So I think that really helps us. Even while burning out in field, those probabilities really help if you know distributions.

[Alli Smith] Awesome. Thank you so much for the explanation. We’ll see what Merlin has to say about this. And Merlin also agrees that it’s a Jungle Babbler and nice work. Everybody gets a point, including the audience. Nice work. All right. So that was a fun bird to start with but let’s see how Merlin does with a bird that’s a little bit more camouflaged.

Here’s question number two. This bird can be found in dry plains and in farmlands but not in forests. We’ll turn the poll on in the audience so you can all vote. And we’ll give everyone a minute or two to vote. All right, the poll is open. In the chat on YouTube, we’ll give everybody about 30 seconds to vote. All right. And we’ll say time is up for voting. We’ll close the poll and we’ll see what the audience had to say here.

All right. It looks like the audience– about 2/3 of the audience voted for Chestnut-bellied Sandgrouse but there were a couple of votes for each of the other species here. Ashwin, what do you think this bird is?

[Samuel Heinrich] So this is a Chestnut-bellied sandgrouse so yes, I think most people have got this right. The structure of this bird, it looks a bit like a pigeon. It’s a little squat. The legs are short. Of course, Francolins also look a bit like this but it’s a little more elongated than a Francolin and this is a Sandgrouse.

Among the two Sandgrouses, again, the distribution helps again here because Chestnut-bellied sandgrouse is distributed through many parts of India but also in parts of Africa, whereas Painted Sandgrouse is subcontinental endemics. So that’s one way to identify this as Chestnut-bellied. Of course, the chestnut belly itself is not the greatest feature but Painted Sandgrouses also more beautifully patterned typically. That also depends on the bill. Yeah.

[Alli Smith] Yeah. Thank you for the description. This is a beautiful bird. And Merlin is agreeing with you that it is a Chestnut-bellied Sandgrouse. And again, nice work. Everybody gets a point here. Everyone got it correct. Nice work.

All right. So this is another great example of a photo that’s perfect for Merlin. Even though this bird is really well camouflaged, it’s a great photo because you can see the whole bird and the bird is in focus and nothing is really obscuring it. But sometimes when you go out and you take photographs of birds, sometimes you don’t get a great photo. Photography is really, really challenging.

Sometimes the bird moves or hides behind a tree or sometimes other birds get in the way. So for this next question, I want you to identify the bird in the background here. Not the blurry bird in the front, the bird that’s in focus in the background. You can see some of its body and some of its bill. So we’ll turn the poll on and we’ll give you all about 30 seconds to vote.

All right. The poll is open. Give everyone a couple seconds to get their votes in here. I know this is a tricky one. I didn’t even show you the whole bird. All right. We will say time is up and we’ll close the poll.

All right. Even though you can’t see the whole bird, most of the audience here voted for Tufted Duck. Everyone’s feeling pretty confident that this might be a Tufted Duck. Ashwin, what do you think about this bird?

[Ashwin Viswanathan] So I agree that it’s a Tufted Duck. This Black and white plumage is quite distinctive and none of the other ducks in the options at least have something similar. But I think there are other clues here as well. For example, the distribution map said it was a migratory duck, which would rule out Indian Spot-billed duck, which is a very commonly seen duck around India and is typically sedentary found locally.

The others, Common Pochard has a reddish head. Otherwise, may look quite similar. At least the males. Females look quite different. So a number of clues that help zone in on Tufted Duck here.

[Alli Smith] Awesome. Let’s see what–

[Ashwin Viswanathan] Yeah.

[Alli Smith] Let’s see what Merlin has to say. Merlin had a creative answer here. Merlin thinks this is a Black headed gull. And it’s not entirely wrong because the bird that’s flying in front of the duck in the background is actually a Black-headed Gull. But the correct answer is Tufted Duck. So Sam, Merlin did not get this one correct. What do you think is going on here?

[Samuel Heinrich] Yeah. So I think there’s a couple of interesting things going on. One is that since Merlin is a machine learning model, you can’t tell it, which bird you want it to identify. You’ve got two on top of each other so it’s going to identify whichever one it feels like.

So in that case, this one was the Black-headed Gull. The other thing is that since this Tufted Duck is a bit obscured, Merlin keys in on coarse grained features a lot of times, which are sort of throughout the entire bird and it does better when it can see the whole bird in focus.

So as you can see here, even if the bird is blurry, Merlin can still do well. But once we start to really sort of block really important parts of the bird for identification, Merlin can get confused and misidentify which species it’s looking at.

[Alli Smith] Thank you for that explanation. That makes sense. Yeah. That is certainly a challenging one, almost a trick question for Merlin but excellent work, Ashwin and the audience for getting that one correct. We’ll move on to our next question, number 4.

I really love birding in late summer and early fall because you get to see young birds like this one here that don’t quite have their adult plumage yet. They’re very cute but they can sometimes be tricky to identify. So what species do you think this is? We’ll turn on the poll so you can vote.

All right. Lots of votes coming in. We’ll give you a few more seconds. All right, time is up. We will close the poll. Thank you all for voting. All right. So the audience here, about 68% voted for Long-tailed Shrike but there are also a lot of votes for some of the other species. Ashwin, what do you think about this one?

[Ashwin Viswanathan] So this is– I can see the questions are getting increasingly more difficult. I think this is a major ID challenge that people face in India, especially in autumn, September, October when many juvenile Shrikes are outside where you see them while birding but you also see a number of migratory Shrikes coming in.

So there can be a lot. It can be quite a difficult time to identify Shrikes. But there are many clues that lead me towards By-backed Shrike in this case, one of them being the size of the bill. It has a very small bill compared to some of the other Shrikes. But again, I’d like to take everybody back to the distribution map.

Long-tailed Shrike, I think many people are aware that it’s also found in the Himalayas. It’s found into Central Asia. It’s found in Northeast India and into Southeast Asia. So Long-tailed Shrike, the distribution alone, it rules it out.

The other Shrikes, which are the other options, Red-tail– Red-tailed is a passage migrant through India. Both Red-tailed and Long-tailed Shrike tend to have slightly larger bills. Even when a bird is a juvenile, I think the structure itself remains what it should be. I’ve forgotten the fourth option. Fourth option was?

[Alli Smith] It was Isabelline, Bay-backed and Red-tailed.

[Ashwin Viswanathan] Isabelline. Right. So Isabelline Shrike I think is the most similar looking among these shrikes because, again, it has a very small bill. And I think the bill is the key here. But Isabelline Shrike, again, is migratory to the Indian subcontinent. And this plumage with a little bit of a reddish back contrasting with the slightly grayish brown head, that you don’t see in Isabelline Shrike so it’s a Bay-backed Shrike.

[Alli Smith] All right. We’ll see what Merlin has to say. Merlin is agreeing with you that it’s a Bay-backed Shrike and nice work. You and Merlin got it correct. I’m so sorry the audience got this one wrong. But Ashwin and Merlin both get a point. Excellent work here and great description for a very tricky bird. We’ll move on to our next question.

So just like young birds can be confusing to identify, so can female birds of some species. The females of a lot of doves in particular can look very similar to one another. So as a clue for this question, this species pictured here is not very common around human activity. What species do you think this is? We’ll turn on the poll.

All right. Everyone can get their votes in to the poll. We’ll give everyone a few more seconds. All right. And time is up for voting. We’ll close the poll and see what the audience has to say.

Ooh, OK. This one, audience is very split on this. Almost equal number of votes for Eurasian Collared-Dove and Red Collared-dove with a couple more votes for Eurasian. Ashwin, what do you think about this bird? This is a tricky one.

[Ashwin Viswanathan] So again, I think this is another very difficult case. And in field, it can be quite tough because often both Eurasian Collared-Dove and Red Collared-Dove occur in the same habitat. And as the audience has said, it is a collared dove. It has a collar. So that’s the first guide we have here.

Beyond that, I think size is extremely indicative. It’s something we can’t see in the photo here but in field, we really need to look out for the size. Red Collared-Dove is visibly smaller than Eurasian Collared-Dove. While they can look extremely, similar female Red Collared-Dove and Eurasian Collared-Dove, size is often a giveaway.

In this case, it’s a female Red Collared-Dove. And this slightly darker plumage Eurasian Collared-Dove tends to be a little grayer. This one, the females also tend to be a little browner. But there’s another very important clue and I keep encouraging everybody to go down that– to look at the distribution. The distribution is often really telling, as is habitat.

Eurasian Collared-Dove tends to be a little more commensal in that it lives with humans a little more easily. Red Collared-Dove as well in some places but typically, in a large part of the subcontinent, it tends to be found a little away from human areas. Eurasian Collared-Dove is found throughout Europe I think all the way into the Americas. It’s invasive, right? In most of the Americas. But the Red Collared-Dove has this distribution in the subcontinent and then into East and Southeast Asia, so.

[Alli Smith] Awesome. Thank you for that beautiful description. We’ll see what Merlin has to say here. Merlin is agreeing with you that this is a Red Collared-Dove. And again, nice work. You and Merlin got this one correct. I’m so sorry. Audience did not get this one correct. It was very, very close in the voting though. But nice work. We’ll give Ashwin and Merlin each a point.

All right. We will move on to our next question. We’re about halfway done with the photo questions now. So here’s our next bird. This one is a very talkative and loud bird that often visits gardens. What species is this? We’ll let you look at the map for a little bit and we’ll turn on the poll in the chat on YouTube to give everybody time to vote.

All right. Lots of votes coming in. We’ll give you a little bit more time. All right. And we’ll say time is up and we’ll close the poll to voting and see what the audience has to say. All right. It looks like the audience is very confident about this one. Almost everybody voted for Himalayan Black Bulbul. Ashwin, what do you think about this?

[Ashwin Viswanathan] So I agree with that. It’s a Himalayan Black Bulbul. Previously, Black Bulbuls, they used to be one species in the Himalayas and Northeast India as well as the Western Ghats but they got split, the Western Ghats one and Square-tailed Bulbul. And I think the distribution alone leads people to Himalayan Black Bulbul.

But I think they were named very well because Square-tailed Bulbul in the Western Ghats is actually blacker than the Himalayan Black Bulbul. But otherwise, just the structure, the crown, the typical Bulbul type crest, the lack of this eye ring, et-cetera rules out Blackbird.

[Alli Smith] Awesome. We will see what Merlin has to say. Merlin is agreeing with everybody. And nice work. Everybody got this answer correct. Excellent. Everyone gets a point. All right. We are going to move back to confusing brown birds for our next question here.

So here’s our next bird. This species can be found throughout most of India in marshes, and grasslands, and farm fields. But there are a lot of similar looking drab brown birds like this. What species is this one in the photo here? We’ll give you time to look at the map and the photo and we’ll turn on the poll in the chat.

All right. Lots of votes coming into the poll. A couple more seconds to get your vote in. All right. And we will say time is up and we’ll close the poll and we’ll see what the audience has to say for this bird. All right. The audience is feeling very confident on this one too. There are a couple votes for each species but almost everybody voted for Plain Prinia. Ashwin, what do you think about this bird?

[Ashwin Viswanathan] I agree. It’s a Plain Prinia. And I think Prinias are a very difficult group of birds. They are brown, they’re nondescript. But I think they have one– they do one thing that makes them I think easier than many other groups in that they vocalize.

So they tend to call a lot. So that is often the best way to identify a Prinia because otherwise, they can look extremely similar. In this case, I think the habitat is also extremely telling. You mentioned it’s found in farmlands. The distribution was throughout the subcontinent and into Northeast India and beyond.

And that is a far too generalist for Delicate Prinia, for example. Delicate Prinia is mostly found in desert areas and along river and grasslands and similar. So that’s one clue that leads us to play in over Delicate.

And other Prinias like Jungle Prinia, et cetera also habitat specialists to an extent. They’re not found in farmlands, not found in open grasslands. So Jungle versus Plain Prinia, a big ID problem in the subcontinent. I think can be resolved through calls as well as where you’re seeing the bird. If you’re seeing it in a farmland or grassland, it’s often not a Jungle Prinia.

[Alli Smith] Awesome. Let’s see what Merlin has to say. And Merlin agrees. And again, everyone gets this one correct. Nice work. Everybody gets a point. So we’ll move on to another tricky category of birds here. There are a lot of very similar looking Warblers like this one on the screen.

And as a clue, this species is an altitudinal migrant, meaning it breeds at high elevations and migrates down to the foothills of mountains during the winter. What species is this? We’ll give you some time to look at the range map on the screen and turn on the poll in the chat.

All right. I see lots of votes coming in. All right. Time is up. We will close the poll. All right. So more divided opinions here in the chat. A lot of votes for all four species here but about half the audience voted for Whistler’s Warbler. Ashwin, what do you think about this bird?

[Ashwin Viswanathan] So I agree that it’s a Whistler’s Warbler. And it’s one of the– it’s from a group of warblers we used to call Golden-Spectacled Warblers. And think that was a really good name because at least in the subcontinent, White-Spectacled Warbler, of course, has a whitish looking spectacle.

But these three birds there, Gray-crowned, Green-crowned and Whistler’s all used to be the same species. They can be extremely similar to look at and now increasingly as people go and explore more and more places in the edges of the country, there are many other similar looking Warblers that are coming into contention. So it’s a bit of a nightmare for ID.

But there’s one easy way to tell apart Gray-crowned, Green-crowned, and Whistler’s here, which is to look at the eye ring. And Gray-crowned and Green-crowned tend to have a broken eye ring. So that way, we can identify this as Whistler’s.

[Alli Smith] Excellent. Let’s see what Merlin has to say. Merlin is agreeing with you all that this is a Whistler’s Warbler. And again, nice work. Everybody got this one correct. So Sam, this bird looks really, really similar to a lot of other species. This one’s a challenge. Sam, how did Merlin handle this? How does Merlin handle similar looking species?

[Samuel Heinrich] Yeah. So before we talked about coarse-grained features, which are the features that are present across an entire bird here, those are kind of useless because these birds look pretty much exactly the same. So Merlin is going to have to key in on the other type of feature it looks at, which is fine grained features.

So it’s going to look at really subtle variations between the species. So Ashwin mentioned the completeness of the eye ring. That’s a great example of a fine grained feature that Merlin is able to key in on and learn that that differentiates Whistler’s Warbler from the other species of Warblers that look quite similar.

[Alli Smith] Awesome. Thank you for that.

[Ashwin Viswanathan] Sorry.

[Alli Smith] Oh, no, no, please go on.

[Ashwin Viswanathan] I’d like to draw everybody’s attention to the White-spectacled Warbler here. This becomes even more difficult because in some other parts of the world, White-spectacled Warbler has a golden spectacle. So I can imagine how confusing this would be for Merlin as well because there are just so much variation even within a single species throughout its range.

[Samuel Heinrich] Certainly. Yeah. And that’s where the size of the Macaulay Library and people contributing their photos. We need tons of photos of species to train Merlin to accurately differentiate them so we can get that variation across their range. So it’s a great point.

[Alli Smith] Yeah. Huge, huge thank you to everybody watching today who submitted their photos to the Macaulay Library. If you’ve ever uploaded your photos to eBird, they come to us into the Macaulay Library and then they’re made available to our machine learning engineers like Sam, who use it to train photo ID. So big, big thank you. We will move on to our next species.

Speaking of birds that can look very similar, here’s our next bird, a type of Nightjar. There are many species that live in India and they’re generally all brown and mottled looking like this. But as a clue, this species can be found in forest clearings and has a really distinctive call that sounds kind of like a cartoon laser gun or something. What species is this? We’ll give you time to look at the range map and we’ll turn the poll on in the chat.

All right. Lots of votes coming in. Excellent. And we’ll say time is up and we’ll close the chat– close the poll for voting and see what the audience has to say about this tricky bird. All right. The audience is kind of split on this one. There are a bunch of votes for each species but over half of the audience voted for Jungle Nightjar. Ashwin, what do you think about this bird?

[Ashwin Viswanathan] I agree with Jungle Nightjar. And I’m can say, I’m glad this photo is during the day. It’s a day roost because identifying Nightjars by plumage is extremely difficult. But what I’m going by mostly here is the fact that it’s roosting on a tree.

And among these species, that’s the only tree rooster or almost a typical tree rooster. Indian Nightjar, Sykes, Savanna, they all tend to roost on the ground typically or just off the ground. And so I think where a Nightjar is sleeping during the day can be quite indicative, also where it is foraging in the night if you’re looking at it.

So Jerdon’s Nightjar is an interesting one. It typically roosts on the ground but it feeds– it perches on raised trees and stumps, et cetera. During the night, Jungle Nightjar tends to roost on trees but it can feet on the ground. It can be up on trees. It’s quite variable in the night. Indian Nightjar is always on the ground. So these are small clues we can use to identify Nightjars.

[Alli Smith] Awesome. Thank you for that explanation of how habitat can really help you identify a bird. Merlin is agreeing with you that this is a jungle Nightjar. And again, everyone gets a point. Nice work audience, Ashwin, and Merlin. All right. We will move on to our next question. We’ve got three questions left for photo ID.

I’ve shown you a lot of really beautiful up close photos of birds but sometimes when you go birding, you only get very distant views of a bird. But even with a distant view, it can still be enough to identify the bird. So here’s a bird that can be found in wetlands but only in the winter. What species is this one? I’ll leave the range map up for a few seconds and we’ll turn on the poll so you can all vote for what you think this species might be.

All right. Lots of votes coming in in the chat. All right. And time is up. We’ll close the poll and we’ll see what the audience has to say. All right. The audience is very confident on this bird. Almost everybody voted for Western Marsh Harrier. Ashwin, what do you think?

[Ashwin Viswanathan] So I think in India, the birding community really loves Raptors and there are so many people who tend to get so very– extremely good at Raptors. And Raptors are fascinating so it’s not surprising.

This is this is a regular Raptor that we see throughout the subcontinent. I agree that it’s a Marsh Harrier, Western Marsh Harrier. It’s a male, which makes it even more distinctive, so. I think this habitat is quite typical as well.

[Alli Smith] Awesome. And we’ll see what Merlin has to say. Merlin is agreeing with you here. And you can really see, especially in the photo that Merlin is showing the distinctive patterning in the wings. And again, nice work. Everybody gets a point. Excellent.

For our next photo, we have another Raptor. This is another Raptor that can be found in the winter but unlike the Harrier, this bird can often be found eating Carrion. What species is this?

Lots of votes coming in to the poll. We’ll give you a few more seconds to vote. All right. And time is up. We’ll see what the audience has to say. Looks like– Ooh, pretty divided here. There are a lot of votes for each of the four species but half the audience voted for Eurasian Griffon. Ashwin, what do you think about this bird? This is an incredible looking bird. They’re so cool.

[Ashwin Viswanathan] Yeah. These vultures are just so lovely. And as we know, they’re becoming increasingly difficult to see. But there are some places in the country where this species can be seen in good numbers but those are also places where it becomes extremely difficult to identify them.

And I think such good views, such good photos are not very common. We often see them in flight way above. So some really important features like the gray legs in this case are not visible. So the other contender is between Eurasian Griffon and Himalayan Griffon and the gray legs and the slightly paler plumage make this a Eurasian Griffon.

[Alli Smith] All right. We’ll see what Merlin has to say. Merlin is agreeing with you. And again, nice work. Everybody got this one correct. Everyone gets a point.

[Ashwin Viswanathan] I’d also like to add that the distribution is, again, extremely helpful here. Where you see the bird or if you have the advantage of looking at the distribution map, you can see that this bird was not distributed through the Himalayas and beyond and instead, it was West into Europe and Africa.

[Alli Smith] Yeah, having that rangemap is a great clue and it might help you with the next question too. So we have our last question coming up. This is our final question in the photo ID round. So sometimes when you’re out birding, you find birds that look a little bit different from others of the same species. Here’s a species that you might find in your garden but one that’s a little unusual looking. What species do you think this is? We’ll turn the poll on in the chat so everybody can vote.

All right. Lots of votes coming in. And time is up for voting. We’ll see what the audience has to say. This is a tricky one but a really, really cool bird. But it looks like the audience is pretty confident that this is a Brahminy Starling as opposed to the other birds. Some of the other species did get votes though. Ashwin, what do you think about this bird here? This really unusual looking one.

[Ashwin Viswanathan] So again, yeah, really, really beautiful bird, this one here. And I would agree with the audience that it’s a Brahminy Starling. And it’s mainly because of– you see the hair coming out here. And an adult– I mean, a normal looking Brahminy Starling, they tend to have this a bit of a crest or a crown, black crest or crown.

And I can imagine if you just change the colors, it’ll look a bit like that. But also the distribution, like you said, is helpful here. It’s resident through most of the subcontinent. I wasn’t aware that it’s a migrant to Sri Lanka. Something I noticed in the map here. And summer visitor a little further North. So that’s quite interesting. But Chestnut-tailed Starling can also be quite similar. It’s migratory to many parts of the Southern Peninsula. So that’s a useful clue.

[Alli Smith] Awesome. We’ll see what Merlin has to say. Merlin thought this was a Chestnut-tailed Starling. That’s kind of interesting. But you are correct. You and the audience got this one right. It is a Brahminy Starling. And Sam, Merlin got this one wrong but it did guess a kind of similar looking species. Could you explain what’s going on here?

[Samuel Heinrich] Yeah. So in the machine learning world, we would call this an out-of-distribution example. So Merlin is trained on millions of photos from the Macaulay Library but it won’t be good at sort of identifying a bird that it’s never seen anything like before. So if it hasn’t been trained on a Leucistic Brahminy Starling, it will have no idea what to do with that.

So it’s able to key in on some of those coarse grain features that are similar across Starlings and it can recognize that it looks like a Starling but not a normal one and so it will guess the one that it thinks is the most likely. And so in this case, Chestnut-tailed Starling has that kind of lightish plumage and Merlin is likely thinking, well, that one looks closer to what I’m seeing so I’m going to go with that.

It doesn’t have the benefit that a human does of knowing what Leucism is and how it affects birds. So it’s a really cool example from a machine learning standpoint. And any time you get to see a Leucistic bird, it’s such a treat. So this is a really fun example.

[Alli Smith] I know. All three the birds on the screen are incredible looking. They’re so cool. And that brings us to the end of the photo ID round. So let’s do a quick score check before we move on. So Merlin has 10 points. Ashwin has 12. Nice work. And the audience has 10. So audience and Merlin are tied right now. Let’s see who’s going to win between those two.

But everyone’s doing great and next up is our sound ID round. So identifying birds by sound can be a really useful skill but it could be really challenging to learn because there are hundreds and hundreds of species and they all make unique sounds.

And one way that we’re working to help you learn is by developing sound ID in Merlin, where Merlin will automatically identify the birds that are singing around you. Sam, could you tell us a bit about how Merlin learns how to identify birds by sound?

[Ashwin Viswanathan] You muted Sam.

[Alli Smith] Oh, Sam. Yep.

[Samuel Heinrich] So thank you. Yes. So similar to photo ID, sound ID starts from the Macaulay Library in a skill group of what we call annotators. So sound recording enthusiasts will upload their recordings to the Macaulay Library and then expert birders will draw boxes around the vocalizations in a spectrogram, which is a visual representation of sound.

And then once we have these annotated recordings with boxes identifying the birds, we will create a machine learning model that learns to identify what spectrogram signal corresponds to what bird.

So yeah, this is a really new sort of technology. It was just recently released in India, thanks to the effort– like countless efforts of sound recordists and annotators working to bring this technology to India. So we’re continuing to hope to expand it. Currently, it works for about 330 species in India and obviously, there’s many more than that in India so we’re continuing to work to try to make sure that we can identify as many of the species as possible.

[Alli Smith] Awesome. Thank you for that explanation. So we’ll have three different questions for sound ID. And for each question, I’ll show you a photo of the habitat that you could find the bird in, and then I’ll play a song from the species twice to give you time to listen.

And we’ll give you to the end of the second playthrough to answer the poll. So we can turn the poll on now before I start playing the sound and we will get started. So this bird can be found in flowering trees in particular. What species might this be?

All right. I will play it again and we’ll get the poll on so you can all vote. All right. We’ll give everyone a second to vote in the poll for what species they think that bird was. And time is up for voting.

We’ll see what the audience has to say. I love the spectrogram on this bird. It’s beautiful the way that the line goes up and down as the bird is singing. All right. Some divided opinions but a lot of the audience voted for Purple-rumped Sunbird. Ashwin, what do you think about this species?

[Ashwin Viswanathan] So just when this song was playing, there was a Spotted Owlet calling outside and I have to remember to report it after this. And it’s amazing that once you start listening to– once you start watching birds, listening to them, you’ll start– even when you’re distracted, you’re listening to other things, there are always bird calls on your mind and I don’t think we miss any really.

And I think just as an aside, we’re planning to hold a nocturnal Bird Count soon so this just reminded me of that. So everyone just look out for that the future. So this, I think there are many clues, the distribution. The habitat, of course, matches everything.

But it’s a very, very common sound we hear, especially in South India and this is a Purple-rumped Sunbird. And like many other birds that eat fruits and flowers, they’ve been doing really well in human areas.

And I think in general, across the country, there’s been a bit of an increase or somewhat stable trajectory for fruit and flower– nectar feeders. So that’s some good news we had from the recent state of India’s birds assessment. So Purple-Sunbird here. Yeah.

[Alli Smith] All right. Also, a nocturnal Bird Count sounds so fun. I’m excited to hear more about how that goes. But let’s see what Merlin has to say about this. Merlin also agrees that this is a Purple-rumped Sunbird and nice work. Everybody gets a point. Excellent.

Good start to the sound ID category. We’ll move on to our next question. Or actually, before we go on, Sam, because sound is such a new thing, this is a pretty familiar sound to a lot of birders. Did Merlin also find this one easy to identify?

[Samuel Heinrich] Yeah. I think since it’s so common and there are so many recordings of it, Merlin has had a lot of chances to learn how to identify this bird. And then I’d also like to point out that the recording we listened to was a very good quality recording. There wasn’t a ton of background noise and you could very clearly hear and see in the spectrogram, the Sunbird calling. So that made this a pretty easy ID for Merlin.

[Alli Smith] Awesome. That makes sense. We will move on to our next question. That was a good warm up but let’s make it a little bit more challenging. So here’s our next bird. Birding by sound in the winter can be kind of challenging because birds aren’t singing as much and instead, are often making much shorter calls and some of these calls can sound very similar.

So here is our next bird. We’ll play it a couple times and get the pole turned on in the chat so you can all vote. But I will play this song twice so you have time to listen. All right. I’ll play it again.

All right. Lots of votes coming in the chat. We’ll give you all a few more seconds to vote. And time is up for voting. Let’s see what everybody has to say. All right. Ooh, divided opinions here. Lots of votes for everything, including a lot of votes for Greenish Warbler. But half the chat voted for Green Warbler. Ashwin, what do you think?

[Ashwin Viswanathan] I agree, Green Warbler. And it has a trisyllabic feel to it. It’s like chi-chewy And I think that’s quite distinctive here. Greenish Warbler tends to– at least the ones that are regularly heard around most of the country tends to have either two or one note so that way, it’s a little easier to tell.

Of course, the distribution map also is quite indicative here. Although the distribution map is due a big update, it’s actually found through a large part of the Peninsula and many places where we thought Green Warbler did not winter previously, the species is actually there. So a very familiar sound in Bangalore city where live in parks, gardens, around home so nice to have this one here.

[Alli Smith] Awesome. Thank you. We’ll see what Merlin has to say. Merlin is agreeing that it’s a Green Warbler. And again, nice work. Everybody got a point here. Everyone got it correct. Sam, here’s another question for you. There are a lot of birds that have short calls like this that sound similar. How does Merlin handle similar sounding birds just like it handles similar looking birds with photo ID?

[Samuel Heinrich] Yeah. So it’ll be a similar concept. So Merlin will need to see a lot of different recordings showing the vocalizations of the species so it can learn to key in on those fine grained features in the spectrogram that differentiate it from other species.

So our current threshold is we hope to have– we need to have hundreds recordings of a bird in order to bring it into the sound ID model. So that gives the model enough time to learn all the variation and precisely how to differentiate a Green Warbler from other Warblers that might have similar calls.

[Ashwin Viswanathan] So I’ll be quite interested to see how Merlin does with the song, which at least I struggle with and I think generally, the human ear struggles a little bit because Green Warbler– some of the Greenish Warblers that come here and many other Greenish Warbler subspecies, they have very similar songs and often quite hard to tell apart, so.

[Samuel Heinrich] Yeah. That’ll be fascinating. These vocal ID challenges are tough for Merlin the same way they’re tough for humans so it’ll be really cool to see.

[Alli Smith] Awesome. Thank you both. We have one more question left for this trivia game. So here, you are birding in a wetland and out in the distance, you hear this noise. Maybe you don’t see the bird but you can hear this coming from out towards the water. We’ll play this song twice and give you all time to vote in the chat.

I love looking at the spectrograms. This one is not nearly as pretty as the others. This is a little bit more of a harsher noise and you can see that represented with that bold, like almost rectangular looking mark on the spectrogram here. I’ll play it one more time for you so you have time to listen and vote.

All right. We’ll get those votes in. And time is up. We’ll close the poll and see what the audience has to say. All right. Ooh, another divided answers here. Lots of votes for both ducks and the Pond-Heron but over half the audience voted for Indian Spot-billed Duck. Ashwin, what do you think?

[Ashwin Viswanathan] So I– to be honest, I don’t know what a Knob-billed duck sounds like. It’s not a very familiar bird to me. I’ve never heard it vocalized. So I hope it’s not a Knob-billed Duck. But think ducks in general, we don’t hear them vocalize very often but it’s very useful to know what they sound like because these circumstances I think while birding, we tend to encounter them often.

I would also guess this is an Indian spot billed duck but this is a great opportunity to be mistaken. All right. Let’s see what Merlin has to say. Merlin didn’t know what this bird was. Merlin has absolutely no idea what it’s listening to. That’s interesting.

And the correct answer is Indian Spot-billed duck. So nice work audience and Ashwin. You both get a point. Merlin does not get a point here. Merlin did not get this one correct. Sam, why is that? This was a pretty clear recording. Why didn’t Merlin get this one?

[Samuel Heinrich] Yeah. So this was super clear, you could hear it really well. But the answer is it’s almost a trick question for Merlin because Merlin has not learned how to identify Indian Spot-billed duck. So there’s a certain number of species that are in the sound ID model and spot billed duck is not one of them yet.

So we have enough recordings of this bird to have it in the model and so what we’re– the process right now is to work with local experts on Indian bird identification to get enough expertly annotated recordings to then onboard the species into the model and make sure that we can identify it well enough to add it into a public model.

So yeah, there’s two steps there. There’s get enough recordings of the bird and then there’s have experts help us precisely identify the bird and then we can release it in Merlin sound ID to the public so that you all can learn how to identify these species and take advantage of the expert knowledge.

[Alli Smith] Awesome. Thank you for that explanation. So hopefully, Indian Spot-billed duck coming to sound sometime in the future after we work on it a little bit more. So thank you, Sam. And that brings us to the end of the trivia game.

So let’s do a final score check. Merlin got 12 points correct. The audience got 13 and Ashwin got 15 out of 15 correct. Nice work, everybody. Audience, you beat Merlin and Ashwin got every single question correct. Ashwin, how are you feeling about this?

[Ashwin Viswanathan] Quite happy. I think it is really nice to go through this and I’d like to generally say that sounds I think are so important for birding and maybe next time, we can have a little more sound when Merlin has more recordings to be trained on. And I think that will be really exciting.

[Alli Smith] Thank you so much for competing today. Before we go, I want to end with a few notes for all of you watching in the audience. Right now, we’re working on making Merlin even better and we need your help to do it.

Merlin is powered by eBird and every eBird checklists you submit helps us make Merlin suggestions even more accurate. And we also want to expand sound ID to include as many species as possible, like that Spot-billed duck. And to do that, we need more sound recordings of the birds from India.

And that means all of you watching in the chat right now, you can all help us. You can use a microphone like the one in the photo here or even your cell phone to go outside and record birds. And if you do record birds and submit those recordings to your eBird checklist, we can use those recordings to train Merlin to identify more species to help you when you’re out in the field.

And a really great place to start is by participating in the Great Backyard Bird Count next week. It runs from February 16th to 19th and all you need to do to participate is watch birds for at least 15 minutes, identify what you saw and submit your sightings to us using Merlin or eBird.

You can learn more at birdcount.org. And last before we go, I want to give a huge thank you to everybody for helping us with this webinar, especially to Bird Count India for making this possible. We have a wonderful team behind the scenes in the chat on YouTube running the polls and answering questions. So thank you to Mittal, Becca, and Chelsea. And a big thank you to Jigyasa Patel from the nature Conservation Foundation for helping design these trivia questions.

And, of course, thank you so much again to our participants tonight. Everyone did a really great job. Thank you so much, Sam for explaining how Merlin is working behind the scenes. Thank you, Ashwin for competing and sharing your tips and tricks and thank you to everybody watching for participating in the polls.

I hope everybody learned something new today and I hope everyone’s looking forward to the Great Backyard Bird Count next week. Thank you all again and have a great night.

[Ashwin Viswanathan] Thank you. Thanks, everyone for participating.

End of transcript

The Great Backyard Bird Count is just around the corner! What better way to brush up on Indian bird identification than with a fun-filled trivia game? The Merlin Bird ID app, developed by the Cornell Lab and powered by machine learning, will be put to the test. Can the app outperform an experienced birder or you, our live audience? Using photos and sounds, we’ll determine who can accurately identify the most bird species found in India.

This is an interactive game, so come ready to play! Throughout the game, we’ll explore how Merlin “decides” on an ID. Our panelist, Ashwin Viswanathan of Bird Count India, will share advice on how to identify tricky birds—plus, he’ll provide insights into birding during the Great Backyard Bird Count.