Smart binoculars can identify 9k birds | Hacker News Reader

Smart binoculars can identify 9k birds

digitalcameraworld.com

158 pointsthunderbong85 comments

Irrespective of other comments, please do support eBird and Merlin, both freely available from the Cornell Lab of Ornithology (CLO).

I am not associated with the CLO other than as a supporter; they do great work and birders come in as many varieties as birds, including me. The flow of data to organizations such as the CLO is critical and these glasses might help augment that.

The data gathering aspect and mapping bird migratory patterns, populations, and other information from AI-enabled field glasses (as well as the apps and other, as there is image recognition in the aforementioned apps and probably many more I'm not aware of) would be quite helpful for a better understanding of how environmental change affects animal populations.

iNaturalist is the other big one - covers all manner of life (not just birds) and much less US focussed. Though not as good as Merlin at ID and birders seem to be split on logging to iNat or eBird - many will log to both

somehow I hope that there is synergy and not destructive, artificial competition between "apps" that do basic nature positive things.. like, does it really need to be "winner take all the AI chat help bots" or something weird like that? hope not

I just compared the 2 apps on the app store. Seek claims the only data used is location and media (images/video). Merlin links identifiers, location, usage data, and contact info.

So, one is either not listing all of the data they harvest from you, or one is a really cool app. I have a feeling both apps are not being fully honest.

They are both run by charities unlike some of the apps in this field. Seek is aimed at children and so collects as little data as possible. It can upload to iNaturalist but only if you connect accounts. Merlin doesn’t collect the data unless you create an account either I don’t believe.

Seek (by iNaturalist) is a good (and fun) app too, it visually identifies all kinds of things (plants, fungi, animals), and integrates with iNaturalist very well.

It works offline, and although not perfect, it is impressive just how much different stuff it has been trained on, and how much you can id even in your own backyard.

> Seek (by iNaturalist) is a good (and fun) app too, it visually identifies all kinds of things (plants, fungi, animals), and integrates with iNaturalist very well.

I left a comment in another thread a while ago complaining that Seek doesn't seem especially concerned with whether its identifications are correct. It was not well received.

I can now substantiate that with species-level identifications provided by Seek, where I have independent knowledge of what I'm looking at. I took several pictures of elephant seals.

For two of them, Seek was willing to make a species-level identification and give me credit for encountering that species.

One was identified as a New Zealand fur seal.

The other was a clouded monitor lizard.

Yes, it's very apparent that it was trained on specific images, e.g. pointing camera at a window insect screen will assume "insect", and a pile of dry leaves is either "mammals" or "tinamous". My orange cat who lays with front paws crossed is a domestic dog (because dogs lay like that). But an aphid colony and specific lichens are identified almost immediately.

In my experience, it's best for small static things, like fungi, lichens, some plants and insects.

Merlin sound ID, too, can produce false identifications (e.g: my relative fooling around was a marsh wren), and I've seen birders frown upon "Identified by Merlin" rare birds posted to eBird.

With Merlin do you need particularly clean loud audio? I've recorded a few sound clips with it: Whilst I can clearly hear at least 4 distinct bird calls in one clip (and recognise a couple), Merlin just doesn't ever recognise any birds at all. I've downloaded a sound pack. I've searched a few times but not found useful help nor even a clip of the app in operation (eg with tips for how to do captures). Any help appreciated, thanks.

Is it better or worse than a human at recognition? Do you need 'perfect' birdsong samples or is there likely something buggy with my install?

Where are you located? It doesn’t support everywhere on the globe yet.

If you’re in a supported area I would suggest just trying it some more. I use it almost daily in the US; I don’t usually save recordings I just tap Sound ID and then cancel it after I get suggestions. It doesn’t always pick up far away birds when they are soft and there is competing noise, and it can get confused by some of the mocking/mimic birds, but it’s generally fantastic at hearing all kinds of sounds despite numerous imperfections and variance in quality.

Thanks for your response; sorry, yes, I'm in UK. From that it sounds like it's not working properly for me; will try it on a different device. Cheers.

none of these work in my area, mockingbirds confuse the hell out of it haha

they register through like 10-15 calls, a lot of birds not even from the region, they get around!

I use the IOS app BirdNET for this - also using the same dataset from Cornell Lab of Ornithology. Highly recommended.

Does anyone know if Merlin Bird Photo ID model has ever been released as open-source?

I am aware of the BirdNET singing ID model being publicly available (the research paper is great for ML and DSP oriented people), but cannot find the same open research for Merlin.

In the UK we have the “BirdTrack” app, which is run by the BTO (British Trust for Ornithology) and also something called iRecord (actually a massive Drupal database…) which covers everything including mammals.

In my view, the problem is these do not solve a real problem in birding.

Binoculars are best when looking with the eyes. Fatigue from holding an object at eye level is a serious issue, so they need to be as light as possible. Swarovski is well known as a first-rate manufacturer of bins.

Cameras with expensive lenses are best for taking photos. With expensive glass and stabilization, they have the "reach" to get enough resolution of the target. Then they can fire off a dozen photos in quick succession and once in a while will capture a genuinely good photo.

Seems to me this product is aimed at people with more money than sense. Too heavy and will take too long to id the bird. There's a reason only gimmicky binoculars come with the ability to take photos.

Yes I don't think experienced birders will go for this. They know what they are looking for and can apply knowledge not just about looks but also about sounds and the bird's jizz. Which means how it behaves, the character of its wing flapping and how it walks on a beach for example. Many of them are well off and buy expensive binocculars, but they also know about tradeoff of weight, light gathering capabilities etc. They want the best "eyes" they can get, and will not trade that for AI.

This product is for rich people that are going to look at birds once in a while. Nothing wrong with that but it is a different demographic.

> also about sounds and the bird's jizz.

I'm reasonably sure that last word is not what you intended to type, but I'm stumped as to what it should be.

macOS dictionary:

    jizz¹ | dʒɪz | vulgar slang
    noun semen
    verb ejaculate
    
    jizz² | dʒɪz | noun, British informal
    (among birdwatchers and naturalists) the characteristic impression given by a particular species of animal or plant: the birds had the familiar jizz of guillemots.

I mean, could be either.

I keep forgetting that this is a field where dickcissel[1], himalayan snowcock[2], or great tit[3] are valid bird names.

[1]: https://en.m.wikipedia.org/wiki/Dickcissel

[2]: https://en.m.wikipedia.org/wiki/Himalayan_snowcock

[3]: https://en.m.wikipedia.org/wiki/Great_tit

What's wrong with great tits?

https://www.printables.com/model/506540-a-great-pair-of-flex...

Fun fact, this is also where the word “jazz” originates: https://en.m.wikipedia.org/wiki/Jazz_(word)

> "Jasm" derives from or is a variant of the slang term "jism" or "gism", which the Historical Dictionary of American Slang dates to 1842 and defines as "spirit; energy; spunk." "Jism" also means semen or sperm, the meaning that predominates today, making "jism" a taboo word.

They mean “GISS” - General Impression, Size, and Shape. Allegedly orginated as an acronym in WWII for identifying planes but caught on in the birding world.

I’d heard the origin among borders as being a short form of “it just is”.

How can you tell that bird’s a blackbird not a starling?

Well, by its.. you know. It just is.

Maybe rizz - bird's mating behavior :)

Well, ok, maybe not.

Actually, it is

https://en.m.wikipedia.org/wiki/Jizz_(birding)

its a real birding term, i have heard older birders say it

Super cool. I love how minimalistic the identification UI is -- just dwell on the subject for a few seconds, and the holographic reticle changes to the name of the bird. No extra text, no complicated interactions. Really nice work making the AI feel integrated and native to the UI.

This piece feels almost like an advertorial, and clearly without any sort of hands on review.

Everything I’ve read about attempts to embed a camera in a pair of binoculars suggests previous results have been poor (given the optics are entirely different).

I’m not a binocular expert, but about the only tech feature I can think of in active use is image stabilisation?

I know nothing about this type of equipment and I’d never heard of the company and thought ‘surely this isn’t the jewellery company?’

It is the same company (different division).

https://en.m.wikipedia.org/wiki/Swarovski

I'm guessing if your grinding jewels or glass at the level they are, there's going to be some technical expertise that is shared.

Does Swarovski cut jewels? I thought all of their stuff was glass.

They do. Lab diamonds to be precise:

https://www.swarovski.com/en-US/c-01/Categories/Jewelry/f/fl...

Also synthetic opal and zirconia etc

Indeed, the same reason Brother makes sewing machines and printers.

And also machine tools, which they originally developed in order to make more and better sewing machines.

https://machinetool.global.brother/

They also make CNC style desktop cutters for things like vinyl/paper

They are actually one of the leading companies when it comes to optics. Swarovski has quite a few interesting business branches, my favourite is SWARCO which makes reflective road markings, traffic management systems and even e-car smart charging solutions.

Birds have an annoying habit of not sitting still, so there is still an opportunity to identify birds by behavior or sounds.

I was birding years ago with two ornithology grad students

We were primarily relying on sound for many but not all species

Australian birds are MUCH louder than European and North American birds

It’s logistically easier to verify species with a high gain omnidirectional field mic than nailing a photo even with a fast AF telescopic lens

We nailed an Albert’s Lyrebird which can only ever realistically be observed via sound, as they’re one of the most evasive birds on the planet when it comes to human interactions

Birdnet can recognize around 3,000:

https://birdnet.cornell.edu/

How can computers learn to recognize birds from sounds? The K. Lisa Yang Center for Conservation Bioacoustics at the Cornell Lab of Ornithology and the Chair of Media Informatics at Chemnitz University of Technology are trying to find an answer to this question. Our research is mainly focused on the detection and classification of avian sounds using machine learning – we want to assist experts and citizen scientist in their work of monitoring and protecting our birds. BirdNET is a research platform that aims at recognizing birds by sound at scale. We support various hardware and operating systems such as Arduino microcontrollers, the Raspberry Pi, smartphones, web browsers, workstation PCs, and even cloud services. BirdNET is a citizen science platform as well as an analysis software for extremely large collections of audio. BirdNET aims to provide innovative tools for conservationists, biologists, and birders alike.

My entry into birdwatching was a superzoom camera and the AI identification tool in iNaturalist. These days I don’t need much tech to ID a bird, but it is what helped me enrich my world in a wonderful way. Smart binos like this just lower the barrier to entry more and I am absolutely thrilled to see it.

I think the price label will be prohibitive for most entrylevel birders. Swarowski bins are either for committed birders that know the value they get and determine it will pay off over years and years of their main hobby, or people for which money is no concern.

I guess we're here: https://xkcd.com/1425/

Probably five minutes with chat GPT vision api

What year was that published?

I'm Pretty sure i first saw tis when I was in university.... around 18 years go

That XKCD was published 2014-09-24. So less than 10 years later, identifying objects in photos has gone from an impossible research task to trivial.

There's a complete list of publication dates at https://www.explainxkcd.com/wiki/index.php/List_of_all_comic...

Well, the comic did say it would take 5 years.

> with a click of a button an identification can be made

I know this is more sensible given the constraints, but I was hoping for a real time HUD…

I rely on eBird but these would be so much better. I joined the Arizona Audubon Society a year ago and I struggle to recognize only about 7 or 8 types of birds on my own. Embarrassing, these smart binoculars would help :-)

Sounds neat but its really expensive and will never get pictures as good as a real camera despite costing as much. You can also just use Merlin on your phone.

> the world’s first AI-supported binoculars

Surely this is only the world's first publicly available AI-supported binoculars. Militaries must have something like this already.

Depends what you mean with "something like this". For example binocculars that uses AI that tags the subject as friend or foe? Something like that is perhaps integrated in advanced weapons systems, but if it was in handheld binocculars, I think it would be known.

All the fly catchers? And ALL non-red tailed hawks?

For those keeping score. There are roughly 9-11k bird species. So these binos claim to be able to ID +90% of them.

After watching reviews: Costs $4700, can't identify small birds, to be released in February.

Would be cool if it could provide suggestions on how to stop he birds from going extinct too.

Spend $4800 vs Download free app... I know which one I would choose...

Sure, but this is pointing to a new vertical for smart objects. The price will be fixed.

Many more tools will become smart, as the phones became smartphones.

I strongly believe the ones that will prevail will be the ones that integrate with phones, people love to control everything from a single place, for example in this case will be binoculars that send the image to the phone and just tell you with voice what did it recognize (or just show it to you if you don't want to make any noise, headphones exist anyway), it has to work offline but there are many models that already can run on the phone, not to mention it's a lot cheaper to delegate processing to the phone, so it's also gonna win price wise, plus its easy to integrate with the services you are used to (e.g. upload to Google photos every time you spot a bird)

The non-smart version of the same binoculars are $3,000. I'd never pay that, but plenty of people do.

https://www.swarovskioptik.com/us/en/birding/products/binocu...

Built-in what???

...ruining a lot of the fun.

Bragging about being able to spot a rare wader on a group of more common but almost identical species is a good chunk of the entire game. If a machine does it for you, the 'aha' moment and the fun searching on the drawings in your book vanish. Is not so interesting anymore and you don't learn the differences. The entire meritocracy based system falls flat.

But can they find our little R2 unit?

Why? What fun is bird watching if you don't even bother to learn about the birds before going into the field? This sounds like going for a "walk" on a Segway. I own some very nice binocs. The one feature that would cause me to throw them off a cliff would be an included "cloud connected app".

I can't speak for everyone, but for me the thrill is in the experience of seeing something, not in rote memorization.

Similarly, when I go out stargazing, I enjoy the beauty of the stars and don't care if I can remember which is Vega and which is Betelgeuse without consulting a reference.

I cannot imagine anyone other than dedicated bird watchers dropping 3000+ on some average-optics binocs.

Again, even for dedicated bird watchers, why is memorizing the details of every kind of bird the important or the fun part?

I'd rather learn from / talk with / have a drink with someone who went out in the field and saw 100 birds without knowing their names than someone who spent their time memorizing the details of 500 and never went in the field.

I don't really understand people who cannot fathom that people can enjoy a hobby in a different manner than they expect, as if their way is the only 'correct' way.

> What fun is bird watching if you don't even bother to learn about the birds before going into the field?

Different people experience hobbies in different ways.

True. The owner of segway scooters also got to experience his hobby in an interesting way.

https://www.cbsnews.com/news/segway-co-owner-drives-segway-o...

When applying for a permit to build a wind turbine, you need to show that there are not too many birds of concern in the area. This kind of technology can make that work cheaper.

I hope they determine that through other means. Putting these binocculars in the hands of a non-expert would not be serious.

Good for discovery and learning. I used to use PictureThis and PictureFish for trees and fish, respectively, so that I could learn what things were. Trying to describe the thing to Google and getting a straight answer is useless - those types of app are magical.

Over time you've learned them all, and don't need it anymore. I could see using binocs like these for the same reason.

Those could be a handy tool for law enforcement the image classification model could be trained to spot criminals? I'm pretty sure there are mugshots from say a sex offender database, then the police just need to take those binoculars to a park… it would be like shooting fish and a barrel!

Identify the jailbirds!

I’ll let that one by, Dad. Just.

I support this joke.

Served time is useless! You’re always guilty!

First thing I thought of. Imagine how the Chinese will use (or perhaps are already using) this technology to track who did what infraction for their Social Credit system. Outside of this, traffic cameras and tracking cars or good, or anything that moves, seems to be a solid use case.

Reply on news.ycombinator.com