AI Distinguishes Birds That Even Experts Can't (sciencemag.org) 32
Slashdot reader sciencehabit quote Science magazine:
It's a fact of life for birders that some species are fiendishly difficult to tell apart — in particular, the sparrows and drab songbirds dubbed "little brown jobs." Distinguishing individuals is nearly impossible. Now, a computer program analyzing photos and videos has accomplished that feat. The advance promises to reveal new information on bird behaviors...
The tool, called a convolutional neural network, sifts through thousands of pictures to figure out which visual features can be used to classify a given image; it then uses that information to classify new images. Convolutional neural networks have already been used to identify various plant and animal species in the wild, including 48 kinds of African animals. They have even achieved a more complicated task for elephants and some primates: distinguishing between individuals of the same species. Team member André Ferreira, a Ph.D. student at the University of Montpellier, fed the neural network several thousand photos of 30 sociable weavers that had already been tagged... [W]hen given photos it hadn't seen before, the neural network correctly identified individual birds 90% of the time, they report this week in Methods in Ecology and Evolution. Behavioral ecologist Claire Doutrelant of CNRS, the French national research agency, says that's about the same accuracy as humans trying to spot color rings with binoculars.
Ferreira then tried the approach on two other bird species studied by Damien Farine, a behavioral ecologist at the Max Planck Institute of Animal Behavior. The tool was just as accurate...
The tool, called a convolutional neural network, sifts through thousands of pictures to figure out which visual features can be used to classify a given image; it then uses that information to classify new images. Convolutional neural networks have already been used to identify various plant and animal species in the wild, including 48 kinds of African animals. They have even achieved a more complicated task for elephants and some primates: distinguishing between individuals of the same species. Team member André Ferreira, a Ph.D. student at the University of Montpellier, fed the neural network several thousand photos of 30 sociable weavers that had already been tagged... [W]hen given photos it hadn't seen before, the neural network correctly identified individual birds 90% of the time, they report this week in Methods in Ecology and Evolution. Behavioral ecologist Claire Doutrelant of CNRS, the French national research agency, says that's about the same accuracy as humans trying to spot color rings with binoculars.
Ferreira then tried the approach on two other bird species studied by Damien Farine, a behavioral ecologist at the Max Planck Institute of Animal Behavior. The tool was just as accurate...
Bird's eye view (Score:5, Interesting)
Feed a trillion readily available pictures into such an algorithm (most uploaded voluntarily to the interwebz) and improve the accuracy a bit, there won't be a bird on the planet that can't be identified individually in any habitat.
I, for one, welcome our bird watching overlords.
Re: (Score:2)
Catching and tagging goes the way of the dodo.
Re: (Score:2)
I doubt it. Tagging gives information about bird migration and community mixing which wouldn't be available from pure counts of species by time and location.
Re: (Score:2)
I doubt it. Tagging gives information about bird migration and community mixing which wouldn't be available from pure counts of species by time and location.
They claim to be able to identify individual birds.
Re: (Score:2)
They claim to be able to identify individual birds.
So unless you can figure out how to capture photos of a particular bird everywhere and everywhen, there is still a use for tagging that provides continual data.
Re: (Score:2)
Yes, but afaik the bulk is don with those small rings.
Re: (Score:2)
There is already such a neural net, Merlin Bird ID, based on about 2 million images from the Macaulay image library.
Re: (Score:2)
Feed a trillion readily available pictures into such an algorithm (most uploaded voluntarily to the interwebz) and improve the accuracy a bit
You mean "Feed a trillion readily available and accurately labelled pictures into such an algorithm". Let me know when you find them - and how you knew you had found them.
Re: Bird's eye view (Score:2)
Or human.
Ba-ba-ba-ba-ba bird is the word...
By the way, you clearly don't know much about birds. Because most of the time, they are sitting behind leaves and such. Only those that survived millions of years of.predators all around them, ... well ... survived, you know? :) ;)
Good luck, though, my shitty trained tensor twiddlers!
Re: (Score:2)
Good luck getting a trillion readily available pictures for training that are tagged to identify individual birds.
Now make it work for audio. (Score:5, Interesting)
Most bird identification is done by listening to the bird sounds. It is great if you can also get a visual id but they can be damn hard to spot. My ornithologist friend is constantly listening to audio tracks of various birds to ensure they can identify the birds they encounter when doing field work.
An interesting project I always wanted to do was to set up multiple audio recording devices positioned ~ 10 m apart from one another. Then when analyzing the audio, one can identify the birds and even triangulate their position. Being able to do this in real time would be amazing and would make an ornithologist's job much easier. They can do it for snipers firing at Humvees (a previous Slashdot posting) so it must also be possible for birds.
Re: (Score:1)
Most bird identification is done by listening to the bird sounds. It is great if you can also get a visual id but they can be damn hard to spot. My ornithologist friend is constantly listening to audio tracks of various birds to ensure they can identify the birds they encounter when doing field work.
Yup. I'm not even a hobbyist when it comes to bird watching, but pretty much every interesting bird I've spotted started with me thinking, "I've never heard that bird before."
Re: (Score:3)
Identifying birds by their songs at least is already available.
https://birdnet.cornell.edu/ [cornell.edu]
For now it's only an android app and not iPhone.
Most of the time it works pretty well.
I found that the noisiest birds in my back yard are the Cardinals and Robins, with a few other things such as woodpeckers.
From ~6 years ago (Score:5, Insightful)
Re:From ~6 years ago (Score:4, Interesting)
oblig XKCD https://xkcd.com/1425/ [xkcd.com]
A lot has changed in six years.
The "virtually impossible" task is now the easier of the two.
You don't even need to train your own CNN. There is a Google Cloud API that can tell you if an image contains a bird in a dozen lines of Python.
Re: From ~6 years ago (Score:3)
That is like saying I coded a browser and the entire internet in a single line of bash: $( firefox ).
Plus, it's in "the cloud". And Google, to add insult to injury. So a complete no-go for any self-respecting developer.
Re: (Score:2)
That is like saying I coded a browser and the entire internet in a single line of bash: $( firefox ).
Any sane solution builds on existing technology.
Do you mine your own iron ore with stone tools and fabricate your own transistors?
Waytogo for that edge-case use : P (Score:1)
Re: Waytogo for that edge-case use : P (Score:2)
Are you implying tensor multiplication gimmics are science?
Call me when the damn thing isn't as close to real neurons as a perfect sphere on a sinusoidal trajectory is to a horse race.
Overlap (Score:2)
Was there overlap between what the humans couldn't identify and the AI? Or did the AI fail ridiculousy when it failed, like when it though a turtle was an AK-47? That's important to know.
Re: Overlap (Score:2)
I'd bet money it was the latter.
Mmmmmh (Score:2)
I distinctly remember scientists having installed cameras in a narrow place to identify penguins as the walk by and they all look the same to me.
Smart camera identifies penguins
14,553 views
Jun 27, 2008
https://www.youtube.com/watch?... [youtube.com]
Justice is finally in sight (Score:3)
I thought I'd never find the culprit that made a mess on my windshield, but now there's finally hope! Your days are numbered, you fluttering shitbird!
I question their differences themselves. (Score:2)
I am fairly certain, that in the majority of groupings, the majority of species have solely been made up, so somebody could get their name in the books.
Like calling humans with one eyebrow that splits like a "Y" (homo eyebrowsplitii monoaeaeiae..ae) a different species from humans with two eyebrows that split like a "Y" (homo eyebrowsplitii vulgaris). Just because you never saw two of them bang.
And don't forget getting mightily offended if somebody suggests they are the same. The sign of any good snob who t
Same problem. (Score:2)
Now train an AI to distinguish Bosons (Score:2)
Why does this matter? (Score:2)
Birds aren’t real!
https://birdsarentreal.com/ [birdsarentreal.com]
Birds Aren't Real (Score:2)
birb (Score:2)
https://www.audubon.org/news/when-bird-birb-extremely-important-guide